View Full Version : Limetech un-Raid Media Storage Server support thread
Pages :
1
[ 2]
3
4
5
6
7
8
limetech 10-08-05, 01:40 AM I looked around for the best price on the Cremax MB123AK drive trays. The best I found was from Census PC. For quantity 12 the price was $22.47 each with shipping. In the end I decided not to go that route primarily because of the cost. Instead I decided to order extra drive cages for the Cooler Master case. Changing out a drive will be a little more difficult, but still doable. One other thing that pushed me away from using a removeable drive tray was the claims I read on usenet about removeable drive trays not conforming to ATA-100 standards. It isn't too often that the cheaper solution gives possibly superior results. If I was using SATA drives I would probably use drive trays.
Scott
There is a wide range of quality among mobile racks, but the Cremax models seem to be a notch above the rest. We have had absolutely no problems using these mobile racks, and believe me, we have evaluated A LOT of mobile racks.
limetech 10-08-05, 01:48 AM ohlwiler,
I have 13 of the Kingwin KF-101 mobile racks in service and have yet to have a drive I/O problem. If anything, I encountered problems using the round IDE cables. I kept getting delayed write errors when using a round IDE cable, the Promise PCI controller card and large (>250Gb) hard drives. Switching to a normal flat 80 wire cable resolved the problem. The CM Stacker case is soooooooooo big that you really do not need to use the round IDE cables as the flat cables will not hinder the air flow inside the case.
Regards,
TCIII
Two comments. First, the Kingwin KF-101 is both a "horrible" mobile rack AND a "great" mobile rack. Horrible because the bottom fan does almost nothing to cool the disk; and worse, after a short amount of time it will actually completely "jam" - that is not rotate at all - as if someone superglued it. BUT, if you remove the fan and remove the top cover, this mobile rack works great because the hole left in the bottom of the tray where the fan was, combined with no top and all-aluminum everywhere else leads to a very cool rack.
Second, you mention a very important point about cables. Do not believe the B.S. about rounded cables being "better" because of better airflow. For one thing, I challenge anyone to actually find a study which shows this. If you talk to guys who design chassis cooling for a living they'll just laugh at the idea of rounded IDE cables. For another thing, you are violating the ATA spec by using rounded cables because the conductors are twisted pairs instead of parallel runs of gound/signal.
So don't waste your money on rounded IDE cables. JMHO.
limetech 10-08-05, 01:57 AM A previous post got me thinking. I'm really not worried about performance with the system, since all my movies and TV shows are xvid encoded. Once I get a video capture card, I'll be storing that locally on my server until I decide it's worth saving. So, the only thing that matters to me is that it would properly calculate the parity data and enough speed to transfer a 1.5gb file in 2 hours. :)
No need to worry about that.
If I find a motherboard that has a Marvell 88E8001 or Intel Pro/1000 ethernet chip, but isn't one of the models you recommend, will it still work? It needs to be able to boot from USB, right? Would it be possible to write the contents of the USB drive to a really small hard drive and boot from that? Would have to wait until you add the feature to exclude drives from the array, correct?
Any m/b you find with a GigE chip on it is not going to be considerably cheaper than the recommended m/b (if at all), so why not just get a recommended m/b? Compared to the cost of the other components and the drives, any difference would be trivial.
If I can find a board that has the Pro/1000 and will take a cheaper processor, I can save a bundle and get it sooner. :)
We use a 2.26GHz Celeron which can be had for 65 bucks. If you manage to get a rebate on just one drive it'll likely pay for most of the processor :)
madpoet 10-08-05, 09:19 AM There are two important characteristics of the m/b:
1. that it supports reliable USB boot, and
2. that it supports GigEthernet.
And one more less important:
3. that it have a separate path to memory for it's GigE chip besides the PCI bus.
Number 1 is the biggest hang up. What m/b are you considering?
THe IC7 Max3. I have it lying around right now. I'd need to confirm the ability to bot from USB though ;). GigE motherboard, Intel 875 based.
http://www.abit-usa.com/products/mb/products.php?categories=1&model=130
Onboard IDE and SATA controllers.
Interesting idea Joe. This is something we can check out, but at first thought I would think it wouldn't speed things up for this reason: typically an OS will buffer blocks read from external storage in it's memory cache. It will then keep track of the occurance of any writes to that buffered copy and mark the block "dirty" if there are any. At some point the OS will need the memory for something else and decide to write out, or flush, the dirty blocks back to storage. Hence, an external storage device, be it a local disk or a remote network share, will probably only "see" write requests for modified blocks.I've never looked at the code dealing with the block buffer cache in a recent version of Linux, but the older System-5 Unix was not smart enough to compare contents of a block currently in the cache with "new" contents being written to that same block. It simply wrote the block to the cache, marked the block as "dirty," (requiring writing to disk) even if it was identical to the existing block and returned. A lower level interrupt driven routine would periodically issue a "sync" command and request the dirty blocks be physically written to the disk.
In the logic I described the "current data" block is read from the disk (and now resides in the buffer cache as "clean") it is then compared to the new contents being written (possibly in the XOR loop) and a flag set if the contents differ.
If no difference, no need to even read the parity block from the parity disk (cluttering up the buffer pool and possibly forcing writes of "dirty" blocks to make room for the parity block being read) since parity will not change for that block. In the same fashion, no need to write the data block marking it as "dirty" as it is not changing.
Now keeping in mind the compare of "old" and "new" data blocks adds overhead when data "is" changing. The question will be is this overhead greater than or less than the potential savings in the overall scheme of things?
Joe L.
madpoet 10-08-05, 12:26 PM Doing my research on the Max3, I can in fact boot from a bootable USB device. So that's another step forward. Now, the Max3 has 2 IDE controllers and 6 SATA controllers. 2 of the SATA controllers are run off the Southbridge chip (and can actually present as IDE drives), and the other 4 are controlled by a Silicon Image chip. This has it's own RAID of course, but it can also (per my understanding) be run without any RAID and should therefore present like any other SATA drive. The question of course is whether your software will have the Silicon Image drivers or something else for your SATA version. I'd be pretty happy though, since if it all worked out I could run 4 IDE drives and 6 SATA drives without even having to add another controller. Always looking to save a buck ;). Anyway, if you've got any clues please let me know. I'll have to wait for your SATA version obviously before I can make any of this work.
Is anyone worried about offsite backup, to protect against losing family photos and other irreplacable media? No matter how good your (un)Raid is, you're still at risk from earthquake, flood, lightning, hurricane, etc.
ohlwiler 10-08-05, 03:02 PM Is anyone worried about offsite backup, to protect against losing family photos and other irreplacable media? No matter how good your (un)Raid is, you're still at risk from earthquake, flood, lightning, hurricane, etc.
I imagine most people that are excited about this product are excited because they have run other solutions. I have had a file server for about 4 years. I have tried several solutions:
Software raid 5: I found the requirement to use the same size of disks to be too restrictive. It was also impossible to expand the array with out first breaking it.
Two big partitions: I have tried spanned disks to create large partitions. I create the main array on one computer and then create a backup array on a second computer. The two arrays are regularly synchronized. Advantages: easy to expand, fast, and fairly safe, two drives have to fail to loose data. Disadvantages: when you loose a drive and have to replace it, it can take a long time to synchronize the data. It also quite inefficient.
When Lime Tech announced their product, I immediately saw the advantages. So what do I store on the file server? I have backups of all the computers in the house. I have media files, audio and video. For my most important data, it also gets backed up to another hard disk and stored at work. Periodically I create backups on optical media of the really important data. What I did was create a list of my different types of data. I decided how important the data was, and what would I do if I lost it. The file server is not my backup solution, it is part of my backup solution.
If I had a fire that burned down my house, I could care less about loosing all of my audio and video files on the file server. All my important documents would be waiting for me at work.
I imagine most here do pretty much the same as I, and most here that are so excited about this product have struggled with different solutions just as I have.
Scott
dougnliz 10-08-05, 03:24 PM I must say I have been trying to find the best solution for a Home Media Server for a while now and this seems to be it. Reliable, easy to expand, and lower cost than most RAID solutions, how can you beat that? The product sounds very exciting and I look forward to getting started. In fact I have all the hardware sitting in my Wish List at NewEgg just waiting for the funds to become available. :D
Before I buy the software though I just wanted to find out if it looks like the Power Buy will happen. I know it has been mentioned as a possibilty and that might just be the shove that pushes me over the edge to get this project going.
Thanks to Lime Tech for coming up with a such a great product and to all of you out there that have posted great questions and feedback. This thread has been extremely educational.
Doug
RADIatiON 10-08-05, 03:56 PM OK..
I just had to jump in and say I'm excited about this also. I have just been starting to look around for a new solution. Currently I have an OpenBSD box with 3 x 250GB Western Digitals with no RAID... a SCSI hooked up to an external Compaq Storage box with 5 x 18GB 10k drives that are software RAID'd as well as three other computers in the house with 100-250GB drives in them (that's a lot of porn!).
This solution looks like the fit that I need to move everything to a central place.
A quick question though on sourcing some components. I feel I could put this beast together myself if I could source all the parts. I live close to Toronto but a quick on-line search has not found me sources for the Cremax cages. The mobo, processor, ram and IDE cards are not a problem.. and the case will cost me a bit to ship... but that's the price you pay for living "way up in the igloo country".
I'm also trying to decide if I should just wait for a SATA solution or not. I have a lot of PATA drives that I could (should) use... but what happens in the future when PATA drives start becoming more expensive than the SATA ones (which is pretty close to happening right now)? Should I just start with some new SATA drives and sell the other beasts, or try to mix-match in the new case?
Anyway... great looking product, and I love the discussion here.
Cheers
Thomas J. Coyle 10-08-05, 03:59 PM dougnliz,
You might want to try zipzoomfly.com as the prices are equal to or less than newegg.com for the motherboard, cpu and ram. Also, they offer free 2nd day shipping where as newegg.com either charges for shipping or only offers 3 day shipping. Just a thought.
Regards,
TCIII
dougnliz 10-08-05, 05:50 PM dougnliz,
You might want to try zipzoomfly.com as the prices are equal to or less than newegg.com for the motherboard, cpu and ram. Also, they offer free 2nd day shipping where as newegg.com either charges for shipping or only offers 3 day shipping. Just a thought.
Regards,
TCIII
Thanks for the tip. It looks like I'll save about $20 just on the items they are offering free shipping on. Most of that is shipping cost. Most of the items offset each other, some are cheaper on ZipZoomFly and some are cheaper on NewEgg, but the free 2nd day shipping makes the deal worth it.
avekevin 10-08-05, 11:56 PM I just ran across this thread, and I have a few questions that I am hoping that someone might anwser.
One of the reasons that RAID-4 is often avoided is that the parity drive becomes the bottleneck for all writes to the system. Every write to every drive also requires an update to the parity drive. How is this overcome?
The unRAID file system sounds a lot like the WAFL approach used by Network Appliance devices. Can you outline the differences?
Thanks,
Kevin
madpoet 10-09-05, 06:51 AM Kevin, they don't avoid it. But once your data is written it is fairly static in a movie/music server, so the read speeds are more than adequate.
loquatsoft 10-09-05, 09:56 AM Which one is better? Price and performance.
Un-Raid vs Enhance Technology QuadraPack
Thanks
Loquat
One of the reasons that RAID-4 is often avoided is that the parity drive becomes the bottleneck for all writes to the system. Every write to every drive also requires an update to the parity drive. How is this overcome?Kevin,
In every raid 4 or 5 system the parity drive requires an update when a data drive is updated. The unRaid is no different.
In Raid 4 (and unRaid) a potential bottleneck exists when *multiple* data drives are written to at the same time and they all need to update their corresponding block on the parity drive "at the same time." Raid-5 splits the parity data up between all the drives and trys to reduce/eliminate the bottleneck by not forcing all the parity writes to the same drive. There is still a potential for drive contention if concurrent writes end up needing to update the same drive that holds their corresponding parity blocks.
I guess what I'm saying is that when writing "one file at a time" the performance of raid-4, raid-5, and unRaid would be roughly similar and no "bottleneck" exists.
In a home media server role, where writing is far less frequent than reading, and where most writing is done "one file at a time" I doubt performance/bottleneck of the parity drive will ever be an issue.
Joe L.
roscoe62 10-09-05, 09:04 PM I'm very interested in this as a media storage solution - I have some doubts about my current solution and have been contemplating options - this one sounds really good! :) , but I'm still interested in taking the DIY path.
Do you have an ETA on when your SATA-only solution will be available? I'm particularly interested in which controller card you plan to use, and also which drivetrays you'll be using for SATA drives.
Cheers!
justmikie 10-09-05, 10:44 PM Tom, here are the replies to your questions:
I'd be interested in knowing some more details about your first test:
a) was ripping done on same PC as viewing?
b) was ripping writing to same disk as the one being read for viewing?
c) what's your network configuration and speed?
d) anything else happening on your network?
a: Ripping was done on a different PC than viewing
b: I tried both. Once played and ripped to disk 2, once played disk 2 and ripped to disk 3. Both times I got stuttering during playback
c: Network switch is gig, I also tried it with a 10/100 switch and got the same amount of stuttering (as near as I cold tell)
d: nothin else goin on....
I was able to do play 2 stream from the same disk (disk2) and it seemed fine, although I only played them both for a minute.
It would be really nice to be able to rip and watch at the same time if you can prioritize the read over the write....
erikatcuse 10-10-05, 02:37 PM Ok
I think I like this unraid the best out of all the options I've looked at but a few questions. Do all the drives need to be the same size? When will a SATA version be ready? What about an AMD version?
I'm thinking when Black friday comes around we should be able to pick up some large HD for cheap. Oh sign me up for a power buy :)
madpoet 10-10-05, 02:57 PM The drives do not need to be the same size, but the parity drive has to be the same size as the largest drive.
The drives do not need to be the same size, but the parity drive has to be the same size as the largest drive.
Actually, the parity drive size has to be equal to or larger than the largest data drive.
Joe L.
Thomas J. Coyle 10-10-05, 03:42 PM gotta5000,
Answer to question #1: No the drives do not have to be the same size, however the parity drive must be equal to or larger than the largest drive.
Answer to question #2: According to the Lime-Technology website, the date is set for 10-17-2005.
Answer to question #3: According to Lime-Technology, the O/S s/w will work on either the Intel motherboard or the Asus A7N8X-E Deluxe motherboard which is an AMD motherboard.
Black Friday has already come around on last Sunday: CompUsa was selling Maxtor 250GB hard drives for $79.99 with no limit.
Regards,
TCIII
madpoet 10-10-05, 03:42 PM Right, sorry ;).
Question, what power supplies are you using since the Stacker doesn't come with any? Just standard 300w?
Thomas J. Coyle 10-10-05, 03:53 PM madpoet,
I am using a pair of Antec 350watt Smart Power Supplies.
Regards,
TCIII
madpoet 10-10-05, 04:05 PM Ok, thanks. Just wanted to make sure there was nothing odd about the requirements.
limetech 10-10-05, 05:51 PM Doing my research on the Max3, I can in fact boot from a bootable USB device. So that's another step forward. Now, the Max3 has 2 IDE controllers and 6 SATA controllers. 2 of the SATA controllers are run off the Southbridge chip (and can actually present as IDE drives), and the other 4 are controlled by a Silicon Image chip. This has it's own RAID of course, but it can also (per my understanding) be run without any RAID and should therefore present like any other SATA drive. The question of course is whether your software will have the Silicon Image drivers or something else for your SATA version. I'd be pretty happy though, since if it all worked out I could run 4 IDE drives and 6 SATA drives without even having to add another controller. Always looking to save a buck ;). Anyway, if you've got any clues please let me know. I'll have to wait for your SATA version obviously before I can make any of this work.
Might not be completely out of the woods yet with USB boot. here are the problems we've encountered with various motherboards:
1. No USB Boot option - bummer.
2. USB Boot option, but "changes" when you change the hard drive configuration - very inconvenient because it means you have to muck with the bios anytime you add or remove a hard drive.
3. USB Boots, but only USB1.1 support - in this case the boot process takes almost a full minute just to read the OS image off the Flash - not ideal, but livable.
4. USB Boots, USB2.0 supported, but once booted OS does not recognize USB. We've seen this with a Asus P4R800V. Solution is to enable only USB1.1.
5. USB Boots, USB2.0 support but still reads slow - this is the case we're finding with Asus A7N8X-E. Once OS is up, USB operates at full speed.
6. USB Boots fast and all is well - this is the case with Intel D865GLCLK.
Regarding SATA on this m/b - I know this is not the answer you want to hear: I just don't know. One thing to check: see if you can disable all SATA support in the bios - this will probably be necessary to boot the current version of the Flash Drive OS.
limetech 10-10-05, 06:17 PM I just ran across this thread, and I have a few questions that I am hoping that someone might anwser.
One of the reasons that RAID-4 is often avoided is that the parity drive becomes the bottleneck for all writes to the system. Every write to every drive also requires an update to the parity drive. How is this overcome?
The unRAID file system sounds a lot like the WAFL approach used by Network Appliance devices. Can you outline the differences?
Thanks,
Kevin
Short answer: the bottleneck is not avoided. Long answer: the UnRAID system is not the right product for applications with heavy random writes. For these systems you are better served using RAID-1 (or RAID-5 if you have to). But, if you want to store alot of media data cheaply and reliably with enough performance for 99% of applications using such data, then UnRAID is for you :D (ok, maybe 99% is a bit high).
Regarding WAFL (Write Anywhere File Layout IIRC?) - This is a good scheme implemented at the file system layer where they are able to ensure only "stripe writes" are the predominate write pattern - in this case RAID-4 is fine. We are not doing this - in UnRAID each disk has it's own self-contained file system just like any disk not in an array.
limetech 10-10-05, 06:28 PM Which one is better? Price and performance.
Un-Raid vs Enhance Technology QuadraPack
Thanks
Loquat
Looking at the QuadraPack it appears this is just a "4-in-3" enclosure; you still have to connect it up to something.
This gives me a good opportunity to talk about 4-in-3, 3-in-2, 5-in-3, etc. Our original intention was to use these enclosures in order to increase the storage density in a case- ie, be able to use a smaller case. But here are the reasons why we ultimately abandoned this idea for the time being:
1. Cost. The QuadraPack above lists for $139. Looking at it from a cost/drive basis that's $34.75 per drive. You can get single drive mobile racks alot cheaper than that.
2. Size. These things tend to be so large that with most small cases you can't get your m/b in without interference, which means you have to use a larger case anyway.
3. Power. Once you get beyond say 8 or 10 drives you really need two power supplies, again pointing to a larger case. [Becaues two 300W power supplies is far less expensive than a single 600W beast.]
So for the Parallel-ATA version of the UnRAID product, cost is major reason why we don't use the higher density enclosures.
For the SATA version, we are again looking at this because there are some enclosures coming on the market that look promising.
limetech 10-10-05, 06:39 PM Right, sorry ;).
Question, what power supplies are you using since the Stacker doesn't come with any? Just standard 300w?
We're using a Sparkle/Fortron FSP300-60PN. This brand has a very good reputation and it's a solid power supply. You can use any high quality p/s. Try to find one with the hightest single rail +12V current rating because that's what hard drives use to spin up with (especially Seagates).
madpoet 10-10-05, 08:07 PM Regarding SATA on this m/b - I know this is not the answer you want to hear: I just don't know. One thing to check: see if you can disable all SATA support in the bios - this will probably be necessary to boot the current version of the Flash Drive OS.
Yes, I can disable all SATA support. It's actually how it's currently set. Good to know that I need to leave it that way ;). I'll just have to try it and see I suppose. I probably have 12 IDE drives lying around anyway.
dougnliz 10-10-05, 08:07 PM Well I'm officially in.
I placed my order for the software today and ordered all the parts from ZipZoomFly and NewEgg. Now I just gotta wait for everything to get here...
I can't wait to get started. I do have a question though. Does the software come with recommended BIOS settings for the Intel MB you recommend?
Thanks,
Doug
limetech 10-10-05, 08:29 PM Well I'm officially in.
I placed my order for the software today and ordered all the parts from ZipZoomFly and NewEgg. Now I just gotta wait for everything to get here...
I can't wait to get started. I do have a question though. Does the software come with recommended BIOS settings for the Intel MB you recommend?
Thanks,
Doug
The bios settings are here. (http://lime-technology.com/bios.htm) (If you see a cat wearing a lime helmet :eek: click Refresh on your browser)
Thomas J. Coyle 10-10-05, 08:41 PM dougnliz,
I assume that you ordered the Intel motherboard from zipzoomfly. I just received my motherboard from them today and unfortunately, you will find no documentation with the motherboard. However, there is a wealth of information about this motherboard on Intel's website: http://support.intel.com/products/motherboard/d865glc/index.htm. Hope this helps.
Regards,
TCIII
dougnliz 10-10-05, 09:39 PM The bios settings are here. (http://lime-technology.com/bios.htm) (If you see a cat wearing a lime helmet :eek: click Refresh on your browser)
Sweet! Thanks for link. I had checked that page earlier but got the cat with the lime helmet.
Nice pic BTW ;)
dougnliz 10-10-05, 09:41 PM dougnliz,
I assume that you ordered the Intel motherboard from zipzoomfly. I just received my motherboard from them today and unfortunately, you will find no documentation with the motherboard. However, there is a wealth of information about this motherboard on Intel's website: http://support.intel.com/products/motherboard/d865glc/index.htm. Hope this helps.
Regards,
TCIII
Thanks TCIII!
Aren't manuals on the web great?!
Thomas J. Coyle 10-11-05, 07:36 PM dougliz,
I did a test set up of the Intel D865GLCLK motherboard, 2.4GHz Celeron D and 512MB of system ram, that I got from zipzoomfly.com yesterday, in a test fixture case that I use to checkout motherboards and peripherals.
Since I usually work with AMD type motherboards, getting the Intel HSF properly aligned on the motherboard was quite an experience. And I must admit that I managed to break one of the cam lever support brackets. However, a little super glue, splints and some patience fixed the cam lever support bracket.
An investigation as to why I managed to break the cam bracket support revealed that there is a very close fit between the motherboard heat sink bracket inner periphery and the fins of the heat sink. Unless you are very careful, you can, like I did, get the heatsink cocked in the heat sink bracket and wind up applying too much force to get the cam lever to close. Therefore make sure that the heatsink fins are not sitting on the edge of the motherboard heatsink bracket and are snugly inside the inner periphery of the motherboard bracket.
If you have the heatsink fins properly seated within the motherboard bracket, it will only take moderate pressure to swing the cam arm to its seated position.
Once I got the motherboard up and running, everything else was a piece of cake. The BIOS correctly detected the processor speed and amount of system memory. Be sure to set up the BIOS per the Lime-Technology instructions. My motherboard is a 409 series (the highest is 413 I believe) and it had the P24 version of the BIOS as required by Lime-Technology.
Lime-Technology,
I let the motherboard run for about an hour and watched the hardware monitoring page in the BIOS. The 2.4 Celeron D with the stock Intel heatsink ramped up to about 113 degress F (44 degrees C) and stayed there. Is this about right or is it on the high side? My Athlon 3200 processors run around this temperature and have been running for a long time.
Regards,
TCIII
dougnliz 10-11-05, 10:08 PM dougliz,
I did a test set up of the Intel D865GLCLK motherboard, 2.4GHz Celeron D and 512MB of system ram, that I got from zipzoomfly.com yesterday, in a test fixture case that I use to checkout motherboards and peripherals.
Since I usually work with AMD type motherboards, getting the Intel HSF properly aligned on the motherboard was quite an experience. And I must admit that I managed to break one of the cam lever support brackets. However, a little super glue, splints and some patience fixed the cam lever support bracket.
An investigation as to why I managed to break the cam bracket support revealed that there is a very close fit between the motherboard heat sink bracket inner periphery and the fins of the heat sink. Unless you are very careful, you can, like I did, get the heatsink cocked in the heat sink bracket and wind up applying too much force to get the cam lever to close. Therefore make sure that the heatsink fins are not sitting on the edge of the motherboard heatsink bracket and are snugly inside the inner periphery of the motherboard bracket.
If you have the heatsink fins properly seated within the motherboard bracket, it will only take moderate pressure to swing the cam arm to its seated position.
Once I got the motherboard up and running, everything else was a piece of cake. The BIOS correctly detected the processor speed and amount of system memory. Be sure to set up the BIOS per the Lime-Technology instructions. My motherboard is a 409 series (the highest is 413 I believe) and it had the P24 version of the BIOS as required by Lime-Technology.
TCIII
Thanks for the tip. It looks like all my hardware will be here on Thursday and my software on Friday. It works out great since I have Friday off and it's supposed to rain all week here. :D
I'll let you know how my build turns out.
Doug
limetech 10-11-05, 11:51 PM dougliz,
I did a test set up of the Intel D865GLCLK motherboard, 2.4GHz Celeron D and 512MB of system ram, that I got from zipzoomfly.com yesterday, in a test fixture case that I use to checkout motherboards and peripherals.
Since I usually work with AMD type motherboards, getting the Intel HSF properly aligned on the motherboard was quite an experience. And I must admit that I managed to break one of the cam lever support brackets....
You're an animal! :D I managed to break a retention bracket once due to the same cause :eek:
I let the motherboard run for about an hour and watched the hardware monitoring page in the BIOS. The 2.4 Celeron D with the stock Intel heatsink ramped up to about 113 degress F (44 degrees C) and stayed there. Is this about right or is it on the high side? My Athlon 3200 processors run around this temperature and have been running for a long time.
Yes, this is within the temperature range we see as well.
Regards,
TCIII[/QUOTE]
limetech 10-11-05, 11:55 PM Thanks for the tip. It looks like all my hardware will be here on Thursday and my software on Friday. It works out great since I have Friday off and it's supposed to rain all week here. :D
I'll let you know how my build turns out.
Doug
Another tip: by default the "rapid bios boot" feature is enabled which means you have to start pressing F2 or Delete immediately after powering-on in order to get into the bios. Later after everytning's configured properly you could re-enable rapid bios boot and shave a few seconds off the boot up time.
limetech 10-12-05, 12:46 AM For those about to build a server using the CM Stacker case, here's the order we follow and some tips.
1. Get your motherboard ready: install CPU, heatsink, and memory. It's also a good idea to hook up a power supply, monitor and keyboard and configure the bios at this time. You don't want to find out something's wrong after it's mounted in the case.
2. Remove the case power supply brackets in order to get at the "goodie" box. First thing to do is instlall the castors to the bottom of the case. This will make everything else go much smoother. Next, continue to strip the case: stock hard disk cage, etc. If you elect to remove the front panel in order to mount 12 drives, do that first. Then it's easy to pop out the plastic bay covers using a flat head screwdriver, working from the top down.
3. Lay case on it's side and install the power supplies and then the motherboard. Use the included adapter to hook up power to the motherboard. The adapter is used to connect the PS_ON signals of the two supplies together. Use cable ties to bundle things neatly. Hook up the case fan power.
4. Set case upright and start installing the mobile racks. The Stacker comes with plastic slides which fit perfectly in the mounting holes of the mobile racks. After putting them on you just slide the mobile rack in until you hear it "click" into place. Start from the bottom and work up. Do not bolt any of the mobile racks in place until they are all in.
5. The top most mobile rack (where the front panel used to be) is the hardest. Use patience - it will go in. Note you have to bolt this one in because you can't use the plastic slides for this one.
6. Bolt all the mobile racks in using one screw on each side. At this point, test each mobile rack tray to ensure it slides in/out easily. Especially the top one.
7. Install the Promise controllers (it doesn't matter which two of the three slots you use) and attach all data cables. Start with M/B Primary to top 2 slots (the end connector to the topmost slot). M/B Secondary to next 2. "IDE1" of PCI controller closest to CPU is next. Then "IDE2" of same controller. Then "IDE1" of controller farthest from CPU; finally, "IDE2 of that one. You will need 4 18" cables and 2 24" cables. The longer cables are necessary for the two "IDE2" connectors to reach their proper mobile racks. Take some time with the cables, by folding them properly you can achieve a very streamlined look.
8. Hook up all your hard drive power. Depending on what supplies you used you may need a couple "Y" splitters. Again, use tie wraps to keep things neat.
9. Ok, you're about done. Install your hard drives, ensuring they're set for Cable Select. Install the Flash and hook up your monitor. Power-on and observe the boot sequence. The first thing you'll see once the Flash starts to boot is the "grub" menu. Grub is a linux menu driven boot loader. You'll have 5 seconds to hit a key to interrupt the boot process, for example to boot an older rev of the software. But let's assume you just let it go. You'll then see linux boot. The process will seem to "pause" when it gets to "hotplug detection". I'm currently using all my spare time to try and figure out how to eliminate this pause - anyway be patient.
10. If you are doing anything different, e.g., installing only one Promise controller, follow the directions included with your Flash drive to edit the "go" script.
11. If you have any problems, email us. If it's really a bugger, we can help over the phone too.
bigshooter 10-12-05, 12:50 AM Have you evaluated any AMD based hardware? I know the intel boards are rock solid, and driver support is great for the integrated components, but a AMD based system may fly through the parity calculations a lot faster than the equivalent Intel based. Also do you see large performance increases as you move to faster processors such as a 3.2ghz prescott on a 800mhz bus vs a 2.26ghz celeron on 533? I know my Athlon 64 3200+ flys though PAR2 calculations faster than my 2.6ghz Pentium 4.
Mark J. Foster 10-12-05, 01:01 AM Hi, Tom!
I've just read through all 15 pages of this thread in one sitting (my head's spinning!), and I'd just like to offer my thanks and congratulations for such a great product, and for such superb support! We AVSers can be a real pain in the a--, being such tightwads and all, but your amazingly patient, open-minded, and thoughtful responses are truly appreciated.
For what it's worth, I'm an "old-timer" that was using PCs before they were even called PCs, have written NetBSD device drivers, have multiple Linux-based commercial product lines in the field, have run a home Linux system for DNS/WWW/MAIL/FTP/SAMBA for over eight years, and have built nearly 50 systems on my own over the years (including 25-CPU and 7-CPU server farms for protein folding, plus four systems that I built from bare PCBs). So... what's the point?
As soon as you're ready to ship the full SATA hardware product, I'm buying the total solution from you. Your incredibly impressive support deserves our support in return. Well done, Tom.
Very Best Wishes!
MarkF
mpenton 10-12-05, 07:51 AM I'd like to add my thanks as well. I received my Flash order Monday and had the server running before dinner. I elected to put the system into a rack mount case (http://www.newegg.com/Product/Product.asp?Item=N82E16811182566) with a 2-to-3 adapter to get the extra drive in. I used an old 1.6 P4 and some noname 266 ram I had laying around. The only changes I'd make would be to put all the documentation you have on the flash. One point though, while it might be somewhere but you don't indicate what the status of your 'fake' leds means. After letting the array construct over-night when I woke up and found all the 'leds' flashing I thought something had gone wrong. Now that I've played with the system I know that it indicates that the hdd is asleep. When I get the array fully loaded I'll do some read rate tests. Right now the system is just slightly slower (subjective, writes only) than the software raid5 system it replaced (dual P3's with individual channels for each drive). Extremely simple to setup ;-)
PS - are you looking at adding LVM support?
ohlwiler 10-12-05, 10:05 AM I got my software form Lime Tech last night. I had previously installed the motherboard, processor, memory, promise cards and drive cages. I am not using removable cages, instead I choose to purchase additional 4 in 3 modules from Cooler Master. At present I am using a single power supply and connected it to the front panel switches. I would like to eliminate the front panel in the future, but I still need to research that a bit. I did look at the "adaptor" that came with the stacker for controlling two power supplies at once. Mine was a 24 pin. Both the motherboard and my power supply use 20 pin connectors. I decided ito postpone hooking it up since mine came with no instructions. Ideally I would just like to flip both power switches on the power supplies like Lime Tech seems to configure their systems. I would appreciate some instructions on how exactly to do this.
I suspected I would have problems with ide cable lengths. I did manage to scrouge around my pile of ide cables and found two 19" cables that have the master and slave connectors a bit closer together. I am opposed to using 24" cables because that exceeds the ata standards by a little too much for my taste. I did experiment with moving a slave connector on a 18" cable so that the master and slave connector are 2" apart. I still need to test this. Since no drives are utilizing the promise cards yet, I will experiment some more.
Setting up the motherboard bios was very easy as Tom posted the instructions just when I needed them. I populated the system with just four drives. I decided to play around with it for a little while to get used to it. A couple of thing confused me. Both are related to the web management utility. Initially on startup I was presented with the option to start the array. I choose to do this. It appeared that nothing was happening. I thought the system would write all zeros to the array. Perhaps it was doing it, but since I had no feedback I could not tell. I decided to click the format button. Unfortunately this part of the management tutorial is vague. The only thing I could find is this snippet at the end of the page: "Note: in the special case where all the hard drives are new, the format operation will not clear the data areas; it simply generates parity." Since I was just experimenting I decided to proceed with the format option. The system told me it would take about 300 minutes to do whatever it was doing. It appeared to skip the clearing of disks and just proceeded to generating parity information. Just for fun I looked to see if it had published the shares to the network. It had, not only that, but I was able to write to them immediately. This had the expected effect of slowing down the parity generation, so I decided to leave it alone until parity generation was complete. After parity generation was complete I started writing to it. Tonight I will try adding another disk to the array.
My suggestion would be to modify the management tutorial with step by step instructions on how to start the array for the first time. This is the time time when the user is the most confused, and needs the most hand holding. My second source of confusion came from the following:
>Initialize, what's that?
This is an option on the Tools page to let you zero-out the array configuration data. The system will think it's brand new with all new hard drives. When you start the array, the system will start a background process to generate the parity information.
The tools tab has no option for Initialize. I suspect there was a wording change somewhere along the road and the instructions and the management utility were not synchronized.
Overall I am simply delighted. It was exactly what I have been searching for. I realized there would be bumps in the road as I am one of the first to try a roll your own system. I suspect that this will be a very successful product. I hope others will post their experiences and impressions to help Tom improve the documentation for the new user.
Scott
Anyone else here feel like they're being tortured by waiting it out for the SATA version?
Good stuff guys, but you're killing me.... :D
Geordon 10-12-05, 12:31 PM Anyone else here feel like they're being tortured by waiting it out for the SATA version?
Good stuff guys, but you're killing me.... :D
Yes, I just received 4 Seagate SATA 300GB drives for about $450 after MIR (see tigerdirect and outpost).
Thomas J. Coyle 10-12-05, 10:29 PM Limetech,
When I had the Intel motherboard setup in my test bed case with nothing except the mouse, keyboard and monitor hooked up, the BIOS would not let me setup a boot drive. I made sure that the Advanced/USB Configuration was correct. The BIOS displayed a USB device in the Boot/HDD menu, but the selection was grayed out.
I assume that when I have the hard drives and the USB Flash Drive attached to the motherboard, I will be able to make the proper selections per your motherboard BIOS Setup instructions.
Regards,
TCIII
zoom8112 10-12-05, 11:36 PM LimeTech,
How did you implement the dual power supplies? Does the CM Stacker have two bays for PSUs?
dougnliz 10-13-05, 01:02 AM I got my software form Lime Tech last night. I had previously installed the motherboard, processor, memory, promise cards and drive cages. I am not using removable cages, instead I choose to purchase additional 4 in 3 modules from Cooler Master. At present I am using a single power supply and connected it to the front panel switches. I would like to eliminate the front panel in the future, but I still need to research that a bit. I did look at the "adaptor" that came with the stacker for controlling two power supplies at once. Mine was a 24 pin. Both the motherboard and my power supply use 20 pin connectors. I decided ito postpone hooking it up since mine came with no instructions. Ideally I would just like to flip both power switches on the power supplies like Lime Tech seems to configure their systems. I would appreciate some instructions on how exactly to do this.
Scott
I have the same question. I bought the Sparkle FSP300-60PN power supplies and they have the 20 pin connector and the separate 4 pin connector. It looks like the recommended Intel MB also has a 20 pin connector and the separate 4 pin one. However, the Stacker came with a single 24 pin connector. How do you make this work?
I also have a question about removing the top panel to put in a 12th hard drive. Since that panel has the power button in it, how will I turn the Stacker on?
Thanks,
Doug
shokunin 10-13-05, 02:51 AM How did you implement the dual power supplies? Does the CM Stacker have two bays for PSUs?
The stacker has dual PSU bays (1 on top, 1 on bottom). It also comes with an adapter to trigger both PSU's to powerup and down.
limetech 10-13-05, 03:50 AM Have you evaluated any AMD based hardware? I know the intel boards are rock solid, and driver support is great for the integrated components, but a AMD based system may fly through the parity calculations a lot faster than the equivalent Intel based. Also do you see large performance increases as you move to faster processors such as a 3.2ghz prescott on a 800mhz bus vs a 2.26ghz celeron on 533? I know my Athlon 64 3200+ flys though PAR2 calculations faster than my 2.6ghz Pentium 4.
At the present time we are concentrating on the SATA version. We have evaluated just one AMD-based m/b (A7n8x-e) & it seems to work o.k., though Flash boot is slower. Once the SATA product is ready we'll return to m/b qualification and other s/w features.
Regarding performance - we haven't tested anything beyond a 2.6Ghz Celeron with a 533MHz bus (the a7n8x-e m/b has an Athlon XP 3000+ "Barton" but with I believe a 400Mhz bus). In our testing so far though, it appears the system is definitely I/O bound and limited primarily by the PCI bus speed and the network.
limetech 10-13-05, 03:51 AM Hi, Tom!
I've just read through all 15 pages of this thread in one sitting (my head's spinning!), and I'd just like to offer my thanks and congratulations for such a great product, and for such superb support! We AVSers can be a real pain in the a--, being such tightwads and all, but your amazingly patient, open-minded, and thoughtful responses are truly appreciated.
For what it's worth, I'm an "old-timer" that was using PCs before they were even called PCs, have written NetBSD device drivers, have multiple Linux-based commercial product lines in the field, have run a home Linux system for DNS/WWW/MAIL/FTP/SAMBA for over eight years, and have built nearly 50 systems on my own over the years (including 25-CPU and 7-CPU server farms for protein folding, plus four systems that I built from bare PCBs). So... what's the point?
As soon as you're ready to ship the full SATA hardware product, I'm buying the total solution from you. Your incredibly impressive support deserves our support in return. Well done, Tom.
Very Best Wishes!
MarkF
Wow, thank you for the vote of confidence :D It is very much appreciated!
limetech 10-13-05, 03:59 AM I'd like to add my thanks as well. I received my Flash order Monday and had the server running before dinner. I elected to put the system into a rack mount case (http://www.newegg.com/Product/Product.asp?Item=N82E16811182566) with a 2-to-3 adapter to get the extra drive in. I used an old 1.6 P4 and some noname 266 ram I had laying around. The only changes I'd make would be to put all the documentation you have on the flash. One point though, while it might be somewhere but you don't indicate what the status of your 'fake' leds means. After letting the array construct over-night when I woke up and found all the 'leds' flashing I thought something had gone wrong. Now that I've played with the system I know that it indicates that the hdd is asleep. When I get the array fully loaded I'll do some read rate tests. Right now the system is just slightly slower (subjective, writes only) than the software raid5 system it replaced (dual P3's with individual channels for each drive). Extremely simple to setup ;-)
PS - are you looking at adding LVM support?
A success story, thanks! That's a great idea adding documentation to the Flash - we will do that. We're updating all the documentation in preparation for the SATA version. I'm glad you figured out the "leds".
LVM support is someting we're considering. I did talk about it a little earlier in the thread. In essence: you give up some features to have LVM: possible to lose all your data if 2 drives die; not possible to upgrade a small disk with a bigger one (unless I suppose it's the one at the "end").
limetech 10-13-05, 04:23 AM I got my software form Lime Tech last night. I had previously installed the motherboard, processor, memory, promise cards and drive cages. I am not using removable cages, instead I choose to purchase additional 4 in 3 modules from Cooler Master. At present I am using a single power supply and connected it to the front panel switches. I would like to eliminate the front panel in the future, but I still need to research that a bit. I did look at the "adaptor" that came with the stacker for controlling two power supplies at once. Mine was a 24 pin. Both the motherboard and my power supply use 20 pin connectors. I decided ito postpone hooking it up since mine came with no instructions. Ideally I would just like to flip both power switches on the power supplies like Lime Tech seems to configure their systems. I would appreciate some instructions on how exactly to do this.
A close look at the 24-pin adapter reveals that the connector that plugs into the motherboard is actually a 20-bin connector "mated" with a 4-pin connector. You can separate the two so that the 20-pin connector can plug into your m/b. The other end of the adapter plugs into the power supply. There will be 4 pins not connected - that's ok, it's the 4 pins on the connector which was separated. Note: this is alot clearer than I'm explaining here.
The adaptor has another power supply connector attached by a pair of very long wires. You plug this into your second power supply so that it recieves the P_ON signal. Finnally, we use tie wraps to hold the connectors together.
I suspected I would have problems with ide cable lengths. I did manage to scrouge around my pile of ide cables and found two 19" cables that have the master and slave connectors a bit closer together. I am opposed to using 24" cables because that exceeds the ata standards by a little too much for my taste. I did experiment with moving a slave connector on a 18" cable so that the master and slave connector are 2" apart. I still need to test this. Since no drives are utilizing the promise cards yet, I will experiment some more.
On some of our systems we've used 20" cables - but these are hard to find. We've also tested 36" cables and didn't have any problems. But strictly speaking, yes, ATA spec is to have no longer than 18" IIRC.
Setting up the motherboard bios was very easy as Tom posted the instructions just when I needed them. I populated the system with just four drives. I decided to play around with it for a little while to get used to it. A couple of thing confused me. Both are related to the web management utility. Initially on startup I was presented with the option to start the array. I choose to do this. It appeared that nothing was happening. I thought the system would write all zeros to the array. Perhaps it was doing it, but since I had no feedback I could not tell. I decided to click the format button. Unfortunately this part of the management tutorial is vague. The only thing I could find is this snippet at the end of the page: "Note: in the special case where all the hard drives are new, the format operation will not clear the data areas; it simply generates parity." Since I was just experimenting I decided to proceed with the format option. The system told me it would take about 300 minutes to do whatever it was doing. It appeared to skip the clearing of disks and just proceeded to generating parity information. Just for fun I looked to see if it had published the shares to the network. It had, not only that, but I was able to write to them immediately. This had the expected effect of slowing down the parity generation, so I decided to leave it alone until parity generation was complete. After parity generation was complete I started writing to it. Tonight I will try adding another disk to the array.
Yep, I apologize for the vagueness of that part of the management utility. We'll try and get that fixed up ASAP.
In a nutshell here's what happened:
a) You installed all new disk drives (that is none had our filesystem on it).
b) When you clicked Start, the array indeed started, but since none of the disks were formatted (that is, had a filesystem), no disk shares were visible - but it did start a background parity genreration (thats the 300 min estimate you saw).
c) The Format button is only present when there exists one or more "unformatted" disks. When you click it, the system will build a ReiserFS file system in those disks. This is analogous to "formatting" a disk in Windows with the NTFS or FAT file system.
d) The format operation is fairly quick, even with parity-sync running in the background. As soon as it finishes, the correspond disk share (s) is exported.
My suggestion would be to modify the management tutorial with step by step instructions on how to start the array for the first time. This is the time time when the user is the most confused, and needs the most hand holding.
Agreed.
My second source of confusion came from the following:
>Initialize, what's that?
This is an option on the Tools page to let you zero-out the array configuration data. The system will think it's brand new with all new hard drives. When you start the array, the system will start a background process to generate the parity information.
The tools tab has no option for Initialize. I suspect there was a wording change somewhere along the road and the instructions and the management utility were not synchronized.
Yes. It's called "Reset array configuration".
Overall I am simply delighted. It was exactly what I have been searching for. I realized there would be bumps in the road as I am one of the first to try a roll your own system. I suspect that this will be a very successful product. I hope others will post their experiences and impressions to help Tom improve the documentation for the new user.
Thank you! Yes there are some rough edges & I appreciate your understanding whle we get through this very busy period.
limetech 10-13-05, 04:26 AM Limetech,
...
I assume that when I have the hard drives and the USB Flash Drive attached to the motherboard, I will be able to make the proper selections per your motherboard BIOS Setup instructions.
Yes that's right. If the m/b doesn't detect any device(s) attached, it won't preesnt anything to choose from in the Boot menu.
limetech 10-13-05, 04:27 AM LimeTech,
How did you implement the dual power supplies? Does the CM Stacker have two bays for PSUs?
Yes.
limetech 10-13-05, 04:34 AM I have the same question. I bought the Sparkle FSP300-60PN power supplies and they have the 20 pin connector and the separate 4 pin connector. It looks like the recommended Intel MB also has a 20 pin connector and the separate 4 pin one. However, the Stacker came with a single 24 pin connector. How do you make this work?
Refer to answer a couple of posts back. If this is not clear I can post a picture.
I also have a question about removing the top panel to put in a 12th hard drive. Since that panel has the power button in it, how will I turn the Stacker on?
You just use the power supply switches, or if they're attached to a power strip or UPS, use that switch.
The key to make this work is the setting in the bios which says what to do when there's a power-fail - be sure it's set to [Power on].
BTW, the proper way to shutdown the system is to Stop the array via the management utility and then just power off. If the system loses power during operation, it will automatically re-boot when power comes back on, and then start a parity check/sync.
dougnliz 10-13-05, 07:19 AM Wow, I must have looked at that adapter 15 times, even looking at one point to see if it separated somehow, and didn't see it. After reading your post it slid right apart. ;)
I just want to make sure I'm clear on this one part. The serparate 4 pin adapter isn't used at all, anywhere from either power supply?
Thanks for the help.
Doug
Thomas J. Coyle 10-13-05, 09:31 AM zoom8112/dougnliz,
I have one of the CM Stackers and it has room for two power supplys. One supply goes in the usual place at the top of the case and the other goes in the bottom. You will have to remove the block off plate from the case at the bottom position and attach the included mounting bracket to the second power supply to mount it. I have the upper power supply supplying the motherboard and four of the hard drives and the lower power supply supplies the other four drives.
What you want to do is take the supplied extender cable and attach it to the upper supply's 20 pin connector. You will notice that the additional four pins on the 20/24 pin connector cable can be detached if your motherboard only uses a 20 pin cable. If your motherboard does not need the 24 pin style connector, detach the extra 4 pin connector and tie it back to the main harness. That 4 pin addition is not to be confused with the normal 4 pin stand alone power supply connector that is used to supply power to the lone 4 pin connector found elsewhere on the Intel motherboard. The 20 pin connector that is attached to the extender cable by the two thin wires is attached to the 20 pin connector on the lower power supply. This way, when the upper power supply starts up, the lower power supply will come on automatically.
Regards,
TCIII
ohlwiler 10-13-05, 09:52 AM I for one am concerned that Tom isn't getting enough sleep. But seriously, thanks for the pointers on setting up the power supplies.
Yesterday I added an additional three drives to my array. This time the software plainly indicated that it would clear the drives prior to adding them to the array. It took about 4 hours to clear them (all 7200, 250GB drives). Next it asked me to format the drives. It then took another 5 hours to recompute parity. Unfortunately I'm moving data from my old file server incrementally so it is time consuming to build the array piecemeal. I wish I had the money to just plunk a whole bunch of new drives in there from the start.
Scott
albrigsr 10-13-05, 11:55 AM ohlwiler,
I , too, had to build my array piecemeal (I bought 2 new 500GB disks) and finally finished using 7 or 8 disks the other night (I can't remember my final count!). Anyway, It's great to reuse the various drives that I had in multiple computers throughout my house and now have everything in a redundant array providing me some much needed protection. I just left 1 drive in each computer for basic files. I originally was going to just put my Video, Music and Picture files on the server but am now thinking of moving most things except basic operating files, etc. to the server. It will make moving to a new computer much easier. I love this product!
Thank you! Yes there are some rough edges & I appreciate your understanding whle we get through this very busy period.
I have a feeling you are going to keep getting busier. ;) Very cool product.
Thomas J. Coyle 10-13-05, 08:35 PM Hi all,
Well, Lime-Technology can claim another satisfied customer!
I assembled the Intel motherboard I tested the other day into my second CH Stacker case along with a 300GB parity drive and seven 250GB data drives all in mobile racks. I decided to leave the front control panel in place and finished the stack with the 4 into 3 sub-chassis at the bottom. Since I presently only have one Promise controller card, I will have to wait to populate the 4 into 3 sub-chassis with four 200GB hard drives that I have as spares.
Just like Tom said, as soon as the BIOS detected the new drives and the OS Flash Drive, I was able to make the Flash Drive the first boot drive. I then hooked up my LAN connection to the motherboard and proceeded to reboot. Checking my router's DHCP page I was able to quickly get the Tower's IP and went right to the webpage. I decided to give the Tower a fixed IP, which was quite simple, since my Linksys router will only let a LAN client have an IP for 24 hrs.
I then started the array and it began to build the parity disk. I followed with the Format command and after a little while each 250GB drive indicated around 245GB of usable disk space.
At first the OS said that it would take approximately 98 hours to build the parity disk, but after a page refresh, the build time was down to approximately 4 hours.
I have decide not to attempt to write to any of the array drives until the parity build is done as I want to run some DVD file transfer rate tests from my number one DVD server to the Tower to get an idea of how long it will take to transfer all of my ripped DVDs. Once I have transferred all of my ripped DVDs, I plan to convert my number one DVD server to the Unraid OS so that both DVD servers will have the same OS.
Regards,
TCIII
dougnliz 10-13-05, 09:36 PM Tip #1 from my install:
If you plan on using the Cremax mobile racks don't buy the optional crossflow fan for the Stacker. It won't fit because the mobile racks are too long.
Doug
Thomas J. Coyle 10-13-05, 09:55 PM dougnliz,
Since I use the Kingwin KF-101 mobile racks, there is no problem with the cross flow fan as the racks do not protrude very far beyond the Stacker hard drive sub-chassis.
With the Kingwin mobile racks, I get drive temperatures of around 37 degrees C with an ambient air temperature of 75 degrees F. Since the DVD server case is in an area where it does not have to be ultra quiet, I was thinking of putting in a higher speed fan for the 120mm fan in the back of the case to improve the air flow. Comments?
Regards,
TCIII
dougnliz 10-13-05, 10:16 PM zoom8112/dougnliz,
I have one of the CM Stackers and it has room for two power supplys. One supply goes in the usual place at the top of the case and the other goes in the bottom. You will have to remove the block off plate from the case at the bottom position and attach the included mounting bracket to the second power supply to mount it. I have the upper power supply supplying the motherboard and four of the hard drives and the lower power supply supplies the other four drives.
What you want to do is take the supplied extender cable and attach it to the upper supply's 20 pin connector. You will notice that the additional four pins on the 20/24 pin connector cable can be detached if your motherboard only uses a 20 pin cable. If your motherboard does not need the 24 pin style connector, detach the extra 4 pin connector and tie it back to the main harness. That 4 pin addition is not to be confused with the normal 4 pin stand alone power supply connector that is used to supply power to the lone 4 pin connector found elsewhere on the Intel motherboard. The 20 pin connector that is attached to the extender cable by the two thin wires is attached to the 20 pin connector on the lower power supply. This way, when the upper power supply starts up, the lower power supply will come on automatically.
Regards,
TCIII
Maybe I'm thick, but I'm still confused on the 4 pin power connector. Here is a snippet from the Intel site concerning the motherboard:
ATX12V power supply
ATX12V compliant power supplies differ from previous ATX power supplies. The ATX12V power supply provides an additional 12V current for the processor via a 2x2 4-pin connector that is required for proper operation.
Now I know I can't use the 4 pin connector with the 24 pin adaptor that Coolermaster supplies, but should I plug the 4 pin connector from the top power supply into the motherboard and just not use the one from the bottom supply?
Thanks for the help.
Doug
Thomas J. Coyle 10-13-05, 11:05 PM Doug,
Yes, you should use the 4 (2X2) pin connector that comes from the upper power supply. The 4 pin connector usually has two yellow and two black wires attached to it. The yellow wires are +12 volts and the black wires are RTN.
The lower power supply should have nothing attached to the 20 pin motherboard connector except for the 20 pin connector with the two wires on it that comes from the extension connector on the upper power supply.
Usually the lower power supply is just used to power the additional drives, case fans and usually nothing more. The upper power supply is usually used to power the motherboard, case fans and the first four or six drives.
Regards,
TCIII
dougnliz 10-13-05, 11:13 PM Doug,
Yes, you should use the 4 (2X2) pin connector that comes from the upper power supply. The 4 pin connector usually has two yellow and two black wires attached to it. The yellow wires are +12 volts and the black wires are RTN.
The lower power supply should have nothing attached to the 20 pin motherboard connector except for the 20 pin connector with the two wires on it that comes from the extension connector on the upper power supply.
Usually the lower power supply is just used to power the additional drives, case fans and usually nothing more. The upper power supply is usually used to power the motherboard, case fans and the first four or six drives.
Regards,
TCIII
Perfect! That's exactly how I have it setup. It actually goes together pretty smooth. I *really* like this case.
Thanks again for the help!
Doug
ohlwiler 10-14-05, 12:42 AM Well I fully populated my system tonight. As I was doing so I also removed a drive that was giving me trouble. Before I added the additional drives I had 7 drives installed. I removed one and added six. When I started the system it warned me that a drive was changed. I went to the tab: tools/reset array configuration. This caused the system to flush the previous configuration leaving my data untouched. Next I went back to the main tab. From here I was given the option to format the six new drives. This is the point that confused me previously, but now I think I understand. The system only zeros out new drives to save having to recompute parity for the array. Since I had to compute parity anyway it skipped clearing my drives prior to formatting. I never saw a clear explanation in the management tutorial of this.
I also added my second power supply, removed the front switch panel and installed the adaptor to allow the motherboard to control both power supplies. It was all very easy. Thanks to all for the explanation on how to do this.
I also played around with some of the other options. The "tools/clear statistics" resets the reads, writes and errors fields on the main tab. The other thing I played with was the comment field in settings/identification. I was running a transfer when I cleared this comment. It appeared to drop the shares off the network to make the change and then published them again. I think I now have a handle on how the system operates. Time to start clamoring for new features! I honestly don't know what else I should ask for though.
Scott
dougnliz 10-14-05, 01:32 AM Woohoo I'm up and running!
I put in my first 2 drives and the system is generating Parity right now. My case is fully setup with the exception of hooking up IDE 2 on each of the Promise Controllers (I need to find some longer IDE cables). Now I just gotta buy more drives! :D
I can't believe how easy this was to setup (once I got past that 4 pin connector thing). :rolleyes: The software was incredibly easy to get going. Great product Tom!!
Thanks to everyone who answered questions for me, especially TCIII and Tom from Lime-Tech! You guys rock!
P.S. One odd thing about my motherboard is that it came with 6 USB ports on the MB instead of 4 like it's supposed to. The one next to the PS/2 connectors has 4 instead of 2. However, the I/O plate only has an opening for 2, so 2 of the ports are behind sheet metal. :confused: Intel must have ran out of the 2 USB port headers...
mpenton 10-14-05, 07:24 AM What temps are all of you running with the removeable racks? I just read where Thomas stated he was running @ 37 degrees with about the same ambient I am. Just running the bare drives installed into my case I'm seeing anywhere from 26-29 degrees. Before I buy the stacker case and the drive racks I want to make sure that I'm buying something that will make the system last longer rather than the opposite. I have mine loaded with 11 250's and last night finished filling the first 7 drives, so far so good. A feature that I really would like to see would be the ability to edit the exported names of the drives. I don't really need them all as one drive but I can already tell that I'm going to have problems remembering what I put where. Yeah, I could write it down but at my age I'm pretty sure that I'd lose the paper ;-) Final question, is there any reason that the flash itself is exported? I'd really rather not expose the flash to the network for any reason at all. I'm already feeling pretty exposed by allowing the drives to be read/write to all users.
dougnliz 10-14-05, 10:21 AM What temps are all of you running with the removeable racks? I just read where Thomas stated he was running @ 37 degrees with about the same ambient I am. Just running the bare drives installed into my case I'm seeing anywhere from 26-29 degrees. Before I buy the stacker case and the drive racks I want to make sure that I'm buying something that will make the system last longer rather than the opposite. I have mine loaded with 11 250's and last night finished filling the first 7 drives, so far so good. A feature that I really would like to see would be the ability to edit the exported names of the drives. I don't really need them all as one drive but I can already tell that I'm going to have problems remembering what I put where. Yeah, I could write it down but at my age I'm pretty sure that I'd lose the paper ;-) Final question, is there any reason that the flash itself is exported? I'd really rather not expose the flash to the network for any reason at all. I'm already feeling pretty exposed by allowing the drives to be read/write to all users.
Well I only have 2 drives in the system right now and I have been copying stuff to the data drive for about 2 hours now and the drives have been holding at 31 and 35 degrees. This morning when they were idle they were at about 25 degrees. I'm using the Cremax mobile racks.
Doug
I was wondering about fan noise / power consumption. When the system is idle and the disks have spun down, do the fans in the mobile trays stop? Are all the fans in the system temperature controlled? Have you measured the power consumption of an idle system?
ohlwiler 10-14-05, 03:14 PM A feature that I really would like to see would be the ability to edit the exported names of the drives.
Now this is a great idea. How about something like:
Slot ......... Export Name
Slot 1 ---> DVDs A-G
Slot 2 ---> DVDs H-R
Slot 3 ---> DVDs R-Z
Slot 4 ---> Music
Slot 5 ---> Photos
Slot 6 ---> Kid's backup
Slot 7 ---> Parent's backup
Alternately you could tie the export name to the model/serial number string. This would allow swapping drives between slots to not matter. When a new drive replaces an old drive model/serial number string is updated already.
Scott
Now this is a great idea. How about something like:
Slot ......... Export Name
Slot 1 ---> DVDs A-G
Slot 2 ---> DVDs H-R
Slot 3 ---> DVDs R-Z
Slot 4 ---> Music
Slot 5 ---> Photos
Slot 6 ---> Kid's backup
Slot 7 ---> Parent's backup
Alternately you could tie the export name to the model/serial number string. This would allow swapping drives between slots to not matter. When a new drive replaces an old drive model/serial number string is updated already.
Scott
I love the idea of tying the export name to the drive model/serial. The admin page could still say drive1, drive2, etc, since the physical drives are involved, but the share names would be much more friendly if they were aliased as you suggested.
Joe L.
boykster 10-14-05, 07:35 PM Regarding higher density enclosures, I have had great luck with these products from servercase.com:
PATA
http://www.servercase.com/miva/miva?/Merchant2/merchant.mv+Screen=PROD&Store_Code=SC&Product_Code=BR-IDE23&Category_Code=ATA100%2F133
SATA
http://www.servercase.com/miva/miva?/Merchant2/merchant.mv+Screen=PROD&Store_Code=SC&Product_Code=BR-SS23&Category_Code=SATA
These are 3 into 2, but they also have 5 into 3. My (non un-Raid) server uses 4 of the 3 into 2 enclosures and they work great. I have them stacked (in a rack case) side by side by side by side and with just the case fans that came with my rack case (2 blow directly on the drive arrays) I have never had any issues with heat issues.
Price is definately higher than the single drive approach, but if space is tight (at least drive bays) then these are great.
Product looks very neat....
Rich
smeehrrr 10-14-05, 07:54 PM I was wondering about fan noise / power consumption. When the system is idle and the disks have spun down, do the fans in the mobile trays stop? Are all the fans in the system temperature controlled? Have you measured the power consumption of an idle system?
Same question here. My servers right now sit in S3 standby most of the time and wake on ping - if Unraid supported the same usage I'd be very tempted to put this together.
albrigsr 10-14-05, 11:38 PM I have 7 drives installed (+1 Parity) and I just took a look at the temperatures after spinning them up for 1 minute and they are as follows:
19, 19, 22, 25, 27, 28, 37 (actually the 37 was already up and running reading data for a while)
Most of the drives typically sit around 20-22 and then move to around 30-35 while working.
My unRaid server arrived on schedule Friday morning and after unpacking it one of the first things I did was take the side panel off. All I can say is that I was very impressed with the worksmanship in its assembly.
I know that most of you will say it pretty much has to only go together one way, but I've seen the insides of a lot of computers and the potential for a mess increases with the number of installed disks. I truly expected there to be a tangled mess when I opened the cabinet.
To my surprise, it was far neater than most any cabinet I've ever had open.
Well done Tom. As I said, I'm impressed.
Joe L.
http://tinypic.com/ekqqzm.jpg
Thomas J. Coyle 10-15-05, 03:09 PM albrigsr,
My warmest running drive during the parity build (the drives were spinning for about five hours straight) was my Seagate 300GB parity drive that reached 40 degree C while the rest of the drives ranged between 34 and 38 degrees C. I think that I will put the parity drive in a different mobile rack that has better cooling capabilities. I have not had a chance to see what the operating temperature is with the drives spun down.
The default spin down time appears to be one hour. Has anyone else set the spin down time to be different than the default value?
Regards,
TCIII
Thomas J. Coyle 10-15-05, 11:28 PM limetech,
After I got my array of eight drives up and running, I decided to add four more drives using the CM Stacker 4 in 3 sub-chassis instead of the Kingwin mobile racks that I have all the other drives in.
When I restarted the array, the webpage indicated that the array was stopped and that four new drives had been detected. I was then given the option to either clear the four new drives only, or, to preserve any data on those four new drives, to just update the parity drive with information from the new drives.
Since I have not added any files to the data drives, and the four new drives were unformatted, I chose to clear the new drives. The new drives are 200GB drives and they are clearing in parallel at the rate of 1% every two minutes. This works out to almost three and one half hours to clear the four drives.
The array Management Tutorial indicates that it should take approximately 45 minutes to clear a 400GB drive. Therefore a 200GB drive should take around 23 minutes. Since my system is clearing the 200GB hard drives at 1% every two minutes, is there something wrong with my system?
Also, the OS will not respond to a webpage update request while clearing the new drives, though the OS is updating the status of the new drive clearing process percent by percent.
Regards,
TCIII
Thomas J. Coyle 10-15-05, 11:39 PM To all,
For those of you who are not using mobile racks, but would like to be able to determine the status of the drives that are attached to the Promise controllers, you will notice that there is a four pin right angle Berg pin connector at the drive end of the board.
According to the Promise user manual, this connector can drive standard drive activity lights to indicate if a drive is being written to or read from. Since I have four drives in the CM Stacker 4 in 3 sub-chassis there is no indication of whether the drives are being accessed as there is with the mobile racks. By using the led drive signal on the Promise controller, drive access status can be made visible using low current leds.
I plan to whip up a little circuit using a Berg pin connector and a couple of leds and test it out. As soon as I can verify proper circuit operation, I will publish the circuit and a parts list.
Regards,
TCIII
albrigsr 10-16-05, 01:27 AM TCIII,
I should have recorded the time to clear my drives since I had to clear 8 of them ranging from 250GB - 500GB. I want to say my 400GB drives took much longer than 45 minutes but I'm not sure though. Maybe 1 - 1.5 hours??? Anyway, I haven't noticed any issues and the speed of writing data to the serverand reading seems great so I think it's just an incorrect benchmark.
TCIII,
I should have recorded the time to clear my drives since I had to clear 8 of them ranging from 250GB - 500GB. I want to say my 400GB drives took much longer than 45 minutes but I'm not sure though. Maybe 1 - 1.5 hours??? Anyway, I haven't noticed any issues and the speed of writing data to the serverand reading seems great so I think it's just an incorrect benchmark.
I just added a pair of 250G Maxtor drives to the unRaid server I purchased from LimeTechnology.
When I restarted the server it prompted me to clear the new drives and warned that it would wipe any existing data. It suggested that I re-initialize the array if I wanted to preserve the data in the new drives instead. Since these were new drives it was not a concern, so I started the clear operation.
The software is in the process of clearing them. At 10 minutes according to the status messages it was about 10% complete. At 16 minutes, it was 15% complete. At 22 minutes it was 20% complete. At 35 minutes it was 30% complete. Looks like it is taking just over one minute per percent.
Everything so far has been very smooth and easy.
Joe L.
curtis104 10-16-05, 04:17 PM Please enlighten me, because I thought all the HDD had to be the same size. Now if I understand correctly, the Parity Drive should be the largest and all the other HDDs' should be of equal value or less to get full capicity use of all the Data HDDs'.
Please enlighten me, because I thought all the HDD had to be the same size. Now if I understand correctly, the Parity Drive should be the largest and all the other HDDs' should be of equal value or less to get full capicity use of all the Data HDDs'.
Your new understanding is correct. Now, in a standard Raid-4/5 the disks do have to be the same size. That is not true in the unRaid server from LimeTechnology. It is what makes it a great option for many of us who can now move the "data" drives from the various PC's in the house to a single central place where the data is protected from most disk crashes. As we move the data we can re-use the drives we currently own and purchase others as we need and sales present themselves. We can mix and match sizes and use all the capacity. We don't need to go out and buy all new equal sized drives.
I started with a 500Gig parity drive and a 500Gig data drive as disk1. So far, I've added two new 250Gig drives, they are disk2 and disk3. Since my Parity drive is a 500Gig drive it will be a long time before I am forced to swap it with one that is larger... and even that is fairly easy to do with the unRaid software. I've still got a 300Gig, a third 250Gig drive, a pair of 120Gig, and an 80Gig drive that will each be relocated to the unRaid server in turn from various PCs in the house. Before I physically move them I will copy their files to the existing unRaid server disks, move them physically, and expand the unRaid array using them. It takes a few hours to physically move and reformat each disk as part of the unRaid array, but the end result will be worth it.
All that will remain in each of my PCs will be a single drive with the OS and the apps used on that specific PC.
Only thing slowing me down is that I only have a 100Megabit lan. Looks like it is time to upgrade the router/NAT... Anybody have any suggestions?
Joe L.
I followed a link from someone's .sig on HTPC news to the Lime site and Google to here :p
Terrific support from Lime - I'm ordering parts left and right from all over the place now. I appreciate the ZipZoomFly linky BTW! Where's the best place to find this case and are the casters included?
The support and responsiveness here is what pushed me over the edge, I had been SERIOUSLY looking at a Terastation but this, this is better IMO and it can grow with me. One Promise card already won on eBay, gathering the rest quickly and you'll have my order soon as well. Your willingness to provide info on hardware that works best rather than trying to turn this into a hardware business with "secret" specs is appreciated! Your openness is another large reason for my purchase, my thanks to you for taking this path. Heh, so much for my stickingto tiny Shuttle XPCs and swearing off full sized hardware! :eek:
I too am interested in SATA together with PATA since the MB appears to support SATA - I currently have none of these in any of my system though but expect to use them in the future. What SATA adapter do you guys recommend? I see that the SATA system is to be released tomorrow - can you give your fans some advanced info? :) You mentioned that the hardware is I/O bound - is there a significant difference in the 133 Promise boards vs 100 in this regard or is it the PCI bus? I actually already have a 66based Promise card, worth even considering? You're using a RAM disk for the OS after it's decompressed and that up to 1Gig has been tested - any reason to go bigger? Mind you if adding additional services I can see this might be needed. I'm not opposed to using an HD for some things too but that's a bit out there for now I'd think. <shrug>
Currently I run a modified Linkstation, it streams data to my Buffalo Linktheater using an onboard uPNP media server. Would it be possible to add such a daemon to this system and if so would you consider it? This would be HUGE for me looking forward since currently I have no HTPC setup (but am looking). Since this is far more open than the stuff on my Linkstation this could really be flexible with a ton of potential IMO.
I too echo the need to have the drives show up as something other than individual share points - a single share point with subdirectories represented by drives would be terrific for the WAF rather than a bunch of additional drive letters. Using the FireWire and USB for still more things would rock - I too am looking for a digital security system (USB cam or capture card) but perhaps that's asking for a bit much in the short term? (lol) Adding security sounds good to me as well but in a family setting it's not so critical I think. :)
Lastly, I guess I'll be upgrading to GigE on my home network! I'm open to suggestions for reasonably priced hardware and good adapter cards that perform well. I've got at least 5 machines but some, like my TIVO, wireless, and WRT54GS, cannot be upgraded so this will certainly be a mixed system for the foreseeable future. :rolleyes:
P.S. One last thing - obviously you guys are using Open source GPL'd software. As I understand it you guys wrote your own proprietary driver which, if I understand GPL correctly, can remain closed source so long as it's not staticly linked in - right? Were any other mods made as I'm sure that eventually someone will ask as you gain popularity and I saw ZERO mention of the GPL browsing your site. Did I miss something or are you comfortable with compliance as it stands? Mind you I'm no rabid Linux fiend so don't get me wrong on this - I generally run Windows actually :eek:
P.P.S. WEB pages on the Lime site coming up without window titles FYI. ;)
bigshooter 10-16-05, 07:05 PM I don't know how many machines you have or what type of network features you need, but you can get an 8 port netgear or linksys 10/100/1000 for anywhere between $70 and $100. As long as your existing router works I'd keep that and just add the switch. I'm going to pick up the 24 port Netgear for $250 unless I can get my Dell rep to throw in a powerconnect 5324 with ou4 next big order of switches (although it will be overkill for my 4 machines :P ).
Mark J. Foster 10-16-05, 08:24 PM Hi, Gang!
If you're looking for an inexpensive Gigabit Ethernet switch, check out Belkin's F5D5141-16 16 Port Gigabit Ethernet switch (http://www.techonweb.com/products/productdetail.aspx?id=A38112&src=FG) for only about $90-95!!! Auto crossover detection on all ports, even!
Cheers!
MarkF
madpoet 10-16-05, 08:39 PM I'm not an expert on this, but it's worth noting that there's been a lot of discussion about GigE and the speeds. You really need to be able to enable jumbo frames to get the significant speed increases, and that requires (I believe) hardware on all your machines able to do it. I could be completely wrong though!
shokunin 10-17-05, 12:40 AM If you go the jumbo frames route and change the MTU all the devices have to have jumbo frames enabled, including your Gigabit switch. Most of the consumer switches don't say or don't support jumbo's. I believe SMC does.
limetech 10-17-05, 01:52 AM I for one am concerned that Tom isn't getting enough sleep. But seriously, thanks for the pointers on setting up the power supplies.
Yesterday I added an additional three drives to my array. This time the software plainly indicated that it would clear the drives prior to adding them to the array. It took about 4 hours to clear them (all 7200, 250GB drives).
I must have been drowsy when I wrote that section about the time it takes to clear the array ;) Obviously the tutorial is wrong: 400GB in 45 min would mean a rate of 148MB/sec, almost 3x faster than the disk! I think what happened is that we often use a special development "hook" in the code to make the disks appear far smaller than they really are for testing purposes & probably I somehow used that number when writing up the tutorial.
Clearing the disks has been a thorn for awhile now. We're currently taking the safest approach; here are the issues:
- When you add a new disk(s), you could always go to the Tools menu and "reset array configuration". The advantage with this is that the disks become available immediately after being formatted (ie, file system built). The disadvantage is that parity is no longer valid and the array will be running "unprotected" until parity-sync completes. So if a drive should happen to fail during this period, you would not be able to reliably rebuild it.
- To avoid having to rebuild parity, if we knew the contents of the new disks were all zeros, then we could just add the disks to the array "as-is" and parity would be correct. Unfortunately, the system has no way of knowing if the disks are already cleared, so that's why we pre-clear. [Actually, the main reason we offer a hard drive product is so that we can test and pre-clear the disk, so when you plug it in the system will know this and skip the pre-clear step.]
- Another unfortunate side-effect of the pre-clear is that the browser won't refresh during this time. We're working on a number of changes to address this issue.
Next it asked me to format the drives.
"Formatting" is the process of creating a file system on the disk - pretty quick.
It then took another 5 hours to recompute parity. Unfortunately I'm moving data from my old file server incrementally so it is time consuming to build the array piecemeal. I wish I had the money to just plunk a whole bunch of new drives in there from the start.
All the shares are exported and you can use the array during the time it's computing parity. Read/write will obviously be a little slower but the system's been tuned to try and minimize the impact of parity-sync on I/O.
limetech 10-17-05, 02:11 AM What temps are all of you running with the removeable racks? I just read where Thomas stated he was running @ 37 degrees with about the same ambient I am. Just running the bare drives installed into my case I'm seeing anywhere from 26-29 degrees. Before I buy the stacker case and the drive racks I want to make sure that I'm buying something that will make the system last longer rather than the opposite. I have mine loaded with 11 250's and last night finished filling the first 7 drives, so far so good.
Regarding disk temperatures - take those with a "grain of salt". The management utility is reporting the value read from the hard drive using S.M.A.R.T. commands. The hard drive manufacturers all say this is not necessarily reliable (in fact I have one Maxtor that reports a temperature of 9C). They also don't say what is the valid range for this sensor, instead they specify the maximum ambient temperature (typically 55C or 60C) and you don't really know where on the drive their sensor is and how accurate it is. What I've found using a probe just under the motor is typically they run in the low-40's when spun up, which is o.k.
A feature that I really would like to see would be the ability to edit the exported names of the drives. I don't really need them all as one drive but I can already tell that I'm going to have problems remembering what I put where. Yeah, I could write it down but at my age I'm pretty sure that I'd lose the paper ;-)
Ok, we'll do it!
Final question, is there any reason that the flash itself is exported? I'd really rather not expose the flash to the network for any reason at all. I'm already feeling pretty exposed by allowing the drives to be read/write to all users.
There's a couple reasons it's exported:
1. To update the software you will have to copy a couple files to the flash. The easiest way to do this is to simply drag them to the Flash share. The next simplest way is to plug the flash into a PC.
2. For test purposes. You can attach the system to your network without any drives installed to test connectivity. In this case you would be able to see at least the Flash share.
But, we will add an option to give you the ability to not export the Flash share.
limetech 10-17-05, 02:14 AM I was wondering about fan noise / power consumption. When the system is idle and the disks have spun down, do the fans in the mobile trays stop? Are all the fans in the system temperature controlled? Have you measured the power consumption of an idle system?
No, no, and no. :rolleyes:
limetech 10-17-05, 02:24 AM I love the idea of tying the export name to the drive model/serial. The admin page could still say drive1, drive2, etc, since the physical drives are involved, but the share names would be much more friendly if they were aliased as you suggested.
Joe L.
Hmm, looking the model/serial number strings, I don't think this would be "more friendly" :)
Our original "share generation" code did this: It exported the names of the top level directories on each disk. So for example, suppose you had this:
disk1:
/Music
/Video
disk2:
/Movies
disk3:
/Movies
disk 4:
/Backup
Then you would see these shares:
Music
Video
Movies.2
Movies.3
Backup
Notice that duplicate directory names generate share names with a numeric suffix corresponding to the disk number. Finally, there's a checkbox that lets you enable or disable exporting the "diskN" shares.
What do you think?
limetech 10-17-05, 02:26 AM Same question here. My servers right now sit in S3 standby most of the time and wake on ping - if Unraid supported the same usage I'd be very tempted to put this together.
It's on the todo list...
limetech 10-17-05, 02:28 AM My unRaid server arrived on schedule Friday morning and after unpacking it one of the first things I did was take the side panel off. All I can say is that I was very impressed with the worksmanship in its assembly.
...
Thanks Joe, it's very much appreciated! We spend a lot of time making sure the system is as solid as possible.
limetech 10-17-05, 02:30 AM TCIII,
I should have recorded the time to clear my drives since I had to clear 8 of them ranging from 250GB - 500GB. I want to say my 400GB drives took much longer than 45 minutes but I'm not sure though. Maybe 1 - 1.5 hours??? Anyway, I haven't noticed any issues and the speed of writing data to the serverand reading seems great so I think it's just an incorrect benchmark.
Yep, incorrect benchmark. See reply a few posts back.
limetech 10-17-05, 03:15 AM I followed a link from someone's .sig on HTPC news to the Lime site and Google to here :p
Terrific support from Lime - I'm ordering parts left and right from all over the place now. I appreciate the ZipZoomFly linky BTW! Where's the best place to find this case and are the casters included?
Yes, they're included.
The support and responsiveness here is what pushed me over the edge, I had been SERIOUSLY looking at a Terastation but this, this is better IMO and it can grow with me. One Promise card already won on eBay, gathering the rest quickly and you'll have my order soon as well. Your willingness to provide info on hardware that works best rather than trying to turn this into a hardware business with "secret" specs is appreciated! Your openness is another large reason for my purchase, my thanks to you for taking this path. Heh, so much for my stickingto tiny Shuttle XPCs and swearing off full sized hardware! :eek:
Once you start generating media files you soon realize that 2, 3, even 4 or more hard drives is just not adequate. Once you get to 6 or 7 it dawns on you how much work it would be to re-generate this if a disk fails. By the time you get to 9 or 10 drives, you've probably suffered a drive failure along the way - that's when you'll be glad you have an UnRAID :D
I too am interested in SATA together with PATA since the MB appears to support SATA - I currently have none of these in any of my system though but expect to use them in the future. What SATA adapter do you guys recommend? I see that the SATA system is to be released tomorrow - can you give your fans some advanced info? :)
See separate post below.
You mentioned that the hardware is I/O bound - is there a significant difference in the 133 Promise boards vs 100 in this regard or is it the PCI bus?
It depends. If you are doing normal read/write, then the bottleneck is definately the hard drives (and the network). If you are generating or checking parity (or clearing new disks), the bottleneck is definately the whole I/O subsystem/memory interface: PCI bus, south bridge IDE controllers, north/south bridge interface, memory bus (so called Front Side Bus), finally CPU. Of the that list, PCI bus is first bottleneck because there are 8 drives on it. Next comes south bridge IDE controllers with 4 drives. On the Intel m/b, the PCI bus and south bridge IDE controllers do seem to operate at least somewhat in parallel, so that helps.
I actually already have a 66based Promise card, worth even considering?
I have never tested that card, but reports from linux newsgroups show that it might be "problematic".
You're using a RAM disk for the OS after it's decompressed and that up to 1Gig has been tested - any reason to go bigger? Mind you if adding additional services I can see this might be needed. I'm not opposed to using an HD for some things too but that's a bit out there for now I'd think. <shrug>
We have to add support to the linux kernel called "high memory" support in order to make use of memory above 1GB. There's code in the driver to suport this as well, but we have not thoroughly tested it. I guess the main reason to have more memory would be to support more services. Somewhere earlier in the thread I think I mentioned that we're not encouraging this because if something goes wrong and the system crashes, you could lose data & our primary role as a "media data server appliance" is to not lose your data :D
Currently I run a modified Linkstation, it streams data to my Buffalo Linktheater using an onboard uPNP media server. Would it be possible to add such a daemon to this system and if so would you consider it? This would be HUGE for me looking forward since currently I have no HTPC setup (but am looking). Since this is far more open than the stuff on my Linkstation this could really be flexible with a ton of potential IMO.
What media server did you have in mind? There's fundamentally no reason why you couldn't do this.
I too echo the need to have the drives show up as something other than individual share points - a single share point with subdirectories represented by drives would be terrific for the WAF rather than a bunch of additional drive letters.
We're looking into this...
Using the FireWire and USB for still more things would rock - I too am looking for a digital security system (USB cam or capture card) but perhaps that's asking for a bit much in the short term? (lol)
Yep.
Adding security sounds good to me as well but in a family setting it's not so critical I think. :)
Agreed, but some kind of simple security is necessary (keyword here is simple).
Lastly, I guess I'll be upgrading to GigE on my home network! I'm open to suggestions for reasonably priced hardware and good adapter cards that perform well. I've got at least 5 machines but some, like my TIVO, wireless, and WRT54GS, cannot be upgraded so this will certainly be a mixed system for the foreseeable future. :rolleyes:
P.S. One last thing - obviously you guys are using Open source GPL'd software. As I understand it you guys wrote your own proprietary driver which, if I understand GPL correctly, can remain closed source so long as it's not staticly linked in - right? Were any other mods made as I'm sure that eventually someone will ask as you gain popularity and I saw ZERO mention of the GPL browsing your site. Did I miss something or are you comfortable with compliance as it stands? Mind you I'm no rabid Linux fiend so don't get me wrong on this - I generally run Windows actually :eek:
Ahhhh, the GPL..... You are correct that we wrote our own driver but it was done in two parts. One part uses derived code from the linux software raid package; the other part is proprietary. The first part we will eventually make public since it's a derived work. The two parts together make a single "module" which is not linked into the kernel, but is loaded at run time. The fact that the two parts themselves are "linked together" I guess is a grey area and I honestly don't know if that's "kosher" or not - if not we can certainly create two modules.
P.P.S. WEB pages on the Lime site coming up without window titles FYI. ;)
I'm not sure what you mean here. I do see window titles using Explorer or Firefox, are you using something else?
limetech 10-17-05, 03:24 AM Here's the update on the SATA code: there's some good news and some bad news.
The good news is that the code works and would be done if it wasn't for the bad news.
The bad news is that some features are not yet supported in the linux SATA subsystem. In particular, support for S.M.A.R.T. is not there, which means:
- no spin down
- no reading temperatures
Word is that "patches" exist that add this support, but they haven't been incorporated into any mainline kernel release. We're actually looking at the code ourselves to see if we can hack this in.
So let me take a quick poll of anyone reading this who's interested: should we release a "pre-release" of the product without this support?
John Spicer 10-17-05, 03:52 AM Drive spin down is a hot spot for me so I'd rather wait to see if it will definitly be included.
Once you have sata will a mix of sata and pata be possibly?
roscoe62 10-17-05, 04:52 AM Drive spin-down is a must-have feature for me also....having said that though, how large of a time-frame are we talking about in order to get those features up & running on SATA? Ballpark?
The reason I ask is that if it's likely to be days or weeks, I can live with that. But if it's MONTHS away, perhaps a "pre-release" without the support is a better option.
Cheers
Thomas J. Coyle 10-17-05, 09:54 AM Tom,
Since your un-Raid product is generating such an enthusiastic response, don't you think that it is about time to either convert this thread to an "Official un-Raid Media Storage Server Thread" or officially start such a thread as a new thread?
Regards,
TCIII
Wow, I appreciate the quick response to my lengthy post!
First and foremost - by all means release somethnig "beta" for SATA. Your upgrade process is pretty simple and this would allow others to help you with the design and testing. Obviously there's some risk involved potentially but many hands make for light work in the testing dept. If the data loss risk was primarily restricted to the SATA drives (?) in a mixed mode then heck I'd buy a drive just to help test. <shrug> SMART temps appear unreliable anyway and spin down while really nice is not a dealbreaker for me since honestly my systems probably don't spin down that much as it is.
Now for some other comments with regards to your responses to me ;)
WEB page titles - the Order Form pages for MB, FOB, and HD are untitled. :rolleyes: It's actually not as bad as I'd first thought but I'm one of those open 2 dozen windows while surfing guys and having window titles helps alot to find the right window. (lol) I've tried tabs too yes.
GPL - please be VERY careful here! DO seperate your modules if at all possible and keep YOUR proprietary code seperate. If you don't, and since you've already distributed code, you could be legally forced to reveal any changes you've made. Seriously, that would suck - just ask Linksys, Sveasoft, and many others. You are also likely legally obligated to post a copy of the GPL on your site and to respond to any requests for modified source of GPL code. If you guys improve upon public GPL code, say the SATA drivers for instance, you've GOT to release it upon request if it's distributed - and besides that helps the community too. That doesn't mean you have to post it for DL but it has to be made "available" (as I understand the GPL). If you statically link into a GPL DLL I believe you'll be hosed :( The BSD code uses a different license and MIGHT be worth looking at if you're in a grey area. You REALLY don't want to end up on Slashdot as the next poster child for the enforcement police although ending up there as an example of cool code seems doable. (lol) If you think there's grey here stay out of it, Linux folk can be pretty rabid about this and for good reason IMO so please don't run afoul of this....
Streaming Media Server... I'm not sure what exactly Linux has to offer. A search on Freshmeat shows alot of hits but many are Beta in nature. Twonkyvision (http://www.twonkyvision.de/) sounds best but isn't free. They do seem to like to build packages for various devices like the NSLU2 (http://www.linuxdevices.com/news/NS7203827694.html), you might be able to work with them and provide an environment they could build to. Their footprint (http://www.twonkyvision.de/UPnP/avsdk.html) appears to be pretty small though and their licensing cost to end users is all of $15. Honestly being placed on their site as a device running their software would probably give you a TON of exposure too :D They specialize in small footprint (http://www.twonkyvision.de/UPnP/index.html) devices for the Mediaserver but honestly if a portion of one of my data disks was needed to support this I'd do it in a heartbeat. BTW note how they are complying with the GPL on their site ;)
For now it sounds like I'll be sticking to just one Gig of memory. It doesn't sound like a NAS would benefit from more but if we begin to add features or services then more might be needed or perhaps sacrifice a portion of a drive. I don't suppose the added memory could be used for caching by chance? Data risky? Speaking of caching - would there be an advantage to using a caching controller card? More expensive I'm sure but might this alleviate some bottlenecks maybe?
Speaking of which you've outlined some of the bottlenecks nicely! Sounds like the cheaper PC100 card it is, I'll leave my 66 cards on the shelf for now. (lol) These were the ones that could be converted to RAID cards but I never got the chance and were purchased for a song. If you'd like I could certainly plug one in as a test though. <shrug>
And yes I fully agree that having some sort of data protection is necessary. I have ripped my CD collection twice now and it looks like it may have been lost again somehow :eek: 100+gigs of MP3 is ALOt of work! I also have a PC that for some reason just loses data, I WAS going to use it for video editing but it's so screwed up I just shut it down. Avoiding weird stuff like that is my goal and so is putting ever larger disks in my PCs with zillions of drive letters. If you can consolidate your share interface I'll do handstands BTW. And as a bonus I may actually, finally, have storage safe enough to back up my home network! I have never lost data, but I've come close. This setup will provide me with piece of mind at last and you had better believe I'll be telling friends about it since everyone I know has handfulls of drives laying around from upgrades. JBOD arrays here we come! :p
ohlwiler 10-17-05, 11:30 AM Our original "share generation" code did this: It exported the names of the top level directories on each disk. So for example, suppose you had this:
disk1:
/Music
/Video
disk2:
/Movies
disk3:
/Movies
disk 4:
/Backup
Then you would see these shares:
Music
Video
Movies.2
Movies.3
Backup
Notice that duplicate directory names generate share names with a numeric suffix corresponding to the disk number. Finally, there's a checkbox that lets you enable or disable exporting the "diskN" shares.
I would be very happy with this solution. It sounds as though you already have it coded up.
Scott
ohlwiler 10-17-05, 11:36 AM Where's the best place to find this case and are the casters included?
I purchased my case from Zipzoomfly. I purchased the Coolermaster STC-T01. It came with casters. It was unclear to me if some of the other Coolermaster variants included casters, so I went with the model I could find reviews on. I also believe someone mentioned obtaining their case from Monarch.
Scott
dahester 10-17-05, 12:47 PM I finally read through this whole thread and must say this is a compelling product! After looking at Terstations and ReadyNASes, this system makes the most sense for large media collections.
What media server did you have in mind? There's fundamentally no reason why you couldn't do this.
There are two popular servers out there I'd like to see built in to the unRAID server. First is mt-daapd (multi-threading digital audio access protocol device), which is an open source iTunes-like audio server.
http://www.mt-daapd.org/
The second is wizd, a compact video server with two main strengths:
1. It can serve HD transport streams to networked DVD players such as the IO Data Avel Linkplayer 2.
2. It can serve DVDs ripped as VIDEO_TS directories.
http://jhue.users.sonic.net/
Even if these servers don't ship with the product, I'd like to see a relatively easy path for the Linux-savvy to modify the flash so these can be added. Thanks.
-Dylan
I finally read through this whole thread and must say this is a compelling product! After looking at Terstations and ReadyNASes, this system makes the most sense for large media collections.
Agreed! I've already begun purchasing peices for mine, still need to order the USB key though - tonight I'd guess :)
There are two popular servers out there I'd like to see built in to the unRAID server. First is mt-daapd (multi-threading digital audio access protocol device), which is an open source iTunes-like audio server.
http://www.mt-daapd.org/
I'd not heard of this one, having only used the one built into my Linkstation streamed to my Linktheater. This one looks like it's geared towards music streamed to iTunes primarily - yes? No video at all near as I can tell but possibly pretty interesting to people. What sort of footprint does it have?
The second is wizd, a compact video server with two main strengths:
1. It can serve HD transport streams to networked DVD players such as the IO Data Avel Linkplayer 2.
2. It can serve DVDs ripped as VIDEO_TS directories.
http://jhue.users.sonic.net/
Even if these servers don't ship with the product, I'd like to see a relatively easy path for the Linux-savvy to modify the flash so these can be added. Thanks.
-Dylan
I REALLY like the idea of streaming HD content if there's a front-end client for this other than dedicated hardware. I'm not HD yet but soon I hope to be and this is attractive. Playing ripped DVDs directly is also VERY attractive! However the URL you've given seems to point to a PPC compiled version for the Kuro\Linkstation (thanks BTW!). The project homepage appears to be here (http://www.geocities.co.jp/SiliconValley-SanJose/3271/) and is Japanese - documented in Kanji from the looks of it :( Not clear on the footprint of this one. Would Twonky suit your needs? It appears to support quite a few formats including VOB and it's claimed footprint is quite small - possibly small enough to keep in memory maybe
madpoet 10-17-05, 03:21 PM God, if you could build in Wizd it would be perfection.
leonowski 10-17-05, 06:28 PM Hi Tom,
Feature Request:
Can we get access to the shares via WebDAV (in addition to SMB)?
dahester 10-17-05, 07:17 PM I REALLY like the idea of streaming HD content if there's a front-end client for this other than dedicated hardware. I'm not HD yet but soon I hope to be and this is attractive. Playing ripped DVDs directly is also VERY attractive! However the URL you've given seems to point to a PPC compiled version for the Kuro\Linkstation (thanks BTW!). The project homepage appears to be here (http://www.geocities.co.jp/SiliconValley-SanJose/3271/) and is Japanese - documented in Kanji from the looks of it :( Not clear on the footprint of this one. Would Twonky suit your needs? It appears to support quite a few formats including VOB and it's claimed footprint is quite small - possibly small enough to keep in memory maybe
Mt-daapd and Wizd are tiny programs (thanks for getting the correct URL on Wizd, BTW). Both are designed to work on embedded 200 MHz PowerPC processors (the Buffalo Linkstation and Linksys NSLU2 are popular devices for creating a NAS with built-in media server). A 2.26 GHz Celery with 512 MB of RAM is plenty of horsepower for the job. The footprint should be small since they have been used successfully on these flash-booted systems.
Mt-daapd is designed to emulate the iTunes server. One special thing about it is it supports on-the-fly transcoding of Apple Lossless audio files to wav, which is perfect for the audiophile freaks out there (like me). It's a very popular server with the Roku Soundbridge crowd, and of course it will work with anything iTunes.
Wizd is special because it has the dual properties of streaming HD transport streams and DVDs with VIDEO_TS file structure intact. It looks like it may be harder to get information from the community at large since it was created in Japan for the popular Avel Linkplayer 2. I believe it will also work with other Sigma Designs-based networked DVD players like the Snazio and Buffalo LinkTheater.
IMO these networked DVD players are the future. Of course, Microsoft and Intel want to be part of that future, but a NAS running Linux that can also serve up your media is the optimal way to bypass their evil plans. :)
madpoet 10-17-05, 07:52 PM There is an English thread for Wizd in the Linkplayer2 forums.
http://www.iodata.com/usa/forum/forumdisplay.php?f=110
There is an English thread for Wizd in the Linkplayer2 forums.
http://www.iodata.com/usa/forum/forumdisplay.php?f=110
This also looks like a good one to look at -> http://www.geocities.com/flipflop7146/wizd.html and best of all it's in English ;)
Honestly, I've got no specific desire for one over another. I would like to be able to do DVDs and I'd like to be able to stream via uPNP to my Buffalo or other front-end. Cheaper is better of course :rolleyes:
Oh yeah - count me as another paying customer as of tonight :)
Geordon 10-17-05, 10:56 PM [QUOTE=BLKMGK]I followed a link from someone's .sig on HTPC news to the Lime site and Google to here :p
Lastly, I guess I'll be upgrading to GigE on my home network! I'm open to suggestions for reasonably priced hardware and good adapter cards that perform well. I've got at least 5 machines but some, like my TIVO, wireless, and WRT54GS, cannot be upgraded so this will certainly be a mixed system for the foreseeable future. :rolleyes:
QUOTE]
I use a Dell PowerConnect 2616 switch. Price was right, but not so sure about the quality. Benchmarks indicated about 3 times faster than 10/100. Not 10x, but still worth the upgrade to me. Did not do anything with jumbo frames, or anything else. For a reliable GigE NIC, I would recommend Intel. I have used the el-cheapo SMC cards with limited success (even with an SMC switch) and would not recommend them.
reel_fan 10-17-05, 11:44 PM Regarding the beta release for SATA...
For me data integrity is priority one. If the data integrity is in question, i'd pass for now.
If it's just the issues listed, and there is a high probability that these features can be added successfully (and tested without compromising the data), I think it would be OK to release the beta.
I work in the software industry, and know all too well how these things usually go ;-)
This may have been asked earlier but I didn't see it, can two drives in the Unraid Server be put into a striped array (raid 0) so that the network will see them as a single drive? I would like to use 300Gb drives in the server, using 2 drives for music and the rest for video. I use a Slimdevices Squeezebox 2 for music playback and I can only use a single folder for playback via the squeezebox. This would require a single drive and I have more than 300 Gb of music burned so far.
Thanks,
Joe M.
boykster 10-18-05, 12:50 AM This may have been asked earlier but I didn't see it, can two drives in the Unraid Server be put into a striped array (raid 0) so that the network will see them as a single drive? I would like to use 300Gb drives in the server, using 2 drives for music and the rest for video. I use a Slimdevices Squeezebox 2 for music playback and I can only use a single folder for playback via the squeezebox. This would require a single drive and I have more than 300 Gb of music burned so far.
Thanks,
Joe M.
Here's a wacky thought for those who want to solve the shares problem....assuming that at the OS level you could have 1 master directory that each of the un-raid drives could be mounted under (with some sort of logical name) and THAT could be the share that is smb exported (samba?). You could still have a drive letter for EACH share, as well as a single unified drive that contains all of the drives are folders....eh, maybe that's how they do what he was describing :
Our original "share generation" code did this: It exported the names of the top level directories on each disk. So for example, suppose you had this:
disk1:
/Music
/Video
disk2:
/Movies
disk3:
/Movies
disk 4:
/Backup
Then you would see these shares:
Music
Video
Movies.2
Movies.3
Backup
But rather than having multiple drives exported, you'd have 1 drive with these (or disk1, disk2, etc) as folders below that master share. Wouldn't require any fancy file system stuff other than mounting the drives systematically under a master folder and exporting that folder.
Rich
limetech 10-18-05, 03:40 AM ...
Once you have sata will a mix of sata and pata be possibly?
Yes, that works.
limetech 10-18-05, 03:45 AM Drive spin-down is a must-have feature for me also....having said that though, how large of a time-frame are we talking about in order to get those features up & running on SATA? Ballpark?
The reason I ask is that if it's likely to be days or weeks, I can live with that. But if it's MONTHS away, perhaps a "pre-release" without the support is a better option.
Cheers
It's a must-have for me too. I'm going to take the conservative estimate and say that it's probably "months" before this support gets into a mainline linux kernel; but, it could happen next week - you never know with open source projects. We are looking into the code ourselves to see if we cobble something together to at least get spindown working.
limetech 10-18-05, 03:46 AM Tom,
Since your un-Raid product is generating such an enthusiastic response, don't you think that it is about time to either convert this thread to an "Official un-Raid Media Storage Server Thread" or officially start such a thread as a new thread?
Regards,
TCIII
I would have to ask permission from the moderators to do that...perhaps.
limetech 10-18-05, 04:05 AM Wow, I appreciate the quick response to my lengthy post!
First and foremost - by all means release somethnig "beta" for SATA. Your upgrade process is pretty simple and this would allow others to help you with the design and testing. Obviously there's some risk involved potentially but many hands make for light work in the testing dept. If the data loss risk was primarily restricted to the SATA drives (?) in a mixed mode then heck I'd buy a drive just to help test. <shrug> SMART temps appear unreliable anyway and spin down while really nice is not a dealbreaker for me since honestly my systems probably don't spin down that much as it is.
I think we will release the code without spindown and temps. There is no issue with data reliability - we've been using SATA-based development systems for months now without problem - of course backed up to an UnRAID server :)
Now for some other comments with regards to your responses to me ;)
WEB page titles - the Order Form pages for MB, FOB, and HD are untitled. :rolleyes: It's actually not as bad as I'd first thought but I'm one of those open 2 dozen windows while surfing guys and having window titles helps alot to find the right window. (lol) I've tried tabs too yes.
Ok I fixed that - nice catch.
GPL - please be VERY careful here! DO seperate your modules if at all possible and keep YOUR proprietary code seperate. If you don't, and since you've already distributed code, you could be legally forced to reveal any changes you've made. Seriously, that would suck - just ask Linksys, Sveasoft, and many others. You are also likely legally obligated to post a copy of the GPL on your site and to respond to any requests for modified source of GPL code. If you guys improve upon public GPL code, say the SATA drivers for instance, you've GOT to release it upon request if it's distributed - and besides that helps the community too. That doesn't mean you have to post it for DL but it has to be made "available" (as I understand the GPL). If you statically link into a GPL DLL I believe you'll be hosed :( The BSD code uses a different license and MIGHT be worth looking at if you're in a grey area. You REALLY don't want to end up on Slashdot as the next poster child for the enforcement police although ending up there as an example of cool code seems doable. (lol) If you think there's grey here stay out of it, Linux folk can be pretty rabid about this and for good reason IMO so please don't run afoul of this...
We will take your advice, thanks!
Streaming Media Server... I'm not sure what exactly Linux has to offer. A search on Freshmeat shows alot of hits but many are Beta in nature. Twonkyvision (http://www.twonkyvision.de/) sounds best but isn't free. They do seem to like to build packages for various devices like the NSLU2 (http://www.linuxdevices.com/news/NS7203827694.html), you might be able to work with them and provide an environment they could build to. Their footprint (http://www.twonkyvision.de/UPnP/avsdk.html) appears to be pretty small though and their licensing cost to end users is all of $15. Honestly being placed on their site as a device running their software would probably give you a TON of exposure too :D They specialize in small footprint (http://www.twonkyvision.de/UPnP/index.html) devices for the Mediaserver but honestly if a portion of one of my data disks was needed to support this I'd do it in a heartbeat. BTW note how they are complying with the GPL on their site ;)
Thanks for the reference, perhaps we'll try and contact those guys.
For now it sounds like I'll be sticking to just one Gig of memory. It doesn't sound like a NAS would benefit from more but if we begin to add features or services then more might be needed or perhaps sacrifice a portion of a drive. I don't suppose the added memory could be used for caching by chance? Data risky? Speaking of caching - would there be an advantage to using a caching controller card? More expensive I'm sure but might this alleviate some bottlenecks maybe?
For someone who wants to get serious about loading other apps on our server, we would really need to provide a way to set aside some swap space. If a process starts to use a lot of memory, or perhaps has a slow memory leak, it would be possible to bring down the system with user data perhaps still cached in memory - not good.
Speaking of which you've outlined some of the bottlenecks nicely! Sounds like the cheaper PC100 card it is, I'll leave my 66 cards on the shelf for now. (lol) These were the ones that could be converted to RAID cards but I never got the chance and were purchased for a song. If you'd like I could certainly plug one in as a test though. <shrug>
Sure go for it; but, I think we need to build-in support for those in the kernel.
And yes I fully agree that having some sort of data protection is necessary. I have ripped my CD collection twice now and it looks like it may have been lost again somehow :eek: 100+gigs of MP3 is ALOt of work! I also have a PC that for some reason just loses data, I WAS going to use it for video editing but it's so screwed up I just shut it down. Avoiding weird stuff like that is my goal and so is putting ever larger disks in my PCs with zillions of drive letters. If you can consolidate your share interface I'll do handstands BTW. And as a bonus I may actually, finally, have storage safe enough to back up my home network! I have never lost data, but I've come close. This setup will provide me with piece of mind at last and you had better believe I'll be telling friends about it since everyone I know has handfulls of drives laying around from upgrades. JBOD arrays here we come! :p
I find I don't map shares to drive letters; I use UNC pathnames instead - but I can see how it would be convenient to consolodate all the storage under one exported share - that will definitely be added to the software.
limetech 10-18-05, 04:11 AM I finally read through this whole thread and must say this is a compelling product! After looking at Terstations and ReadyNASes, this system makes the most sense for large media collections.
There are two popular servers out there I'd like to see built in to the unRAID server. First is mt-daapd (multi-threading digital audio access protocol device), which is an open source iTunes-like audio server.
http://www.mt-daapd.org/
The second is wizd, a compact video server with two main strengths:
1. It can serve HD transport streams to networked DVD players such as the IO Data Avel Linkplayer 2.
2. It can serve DVDs ripped as VIDEO_TS directories.
http://jhue.users.sonic.net/
Even if these servers don't ship with the product, I'd like to see a relatively easy path for the Linux-savvy to modify the flash so these can be added. Thanks.
-Dylan
I looked into wizd at this site:
http://www.geocities.com/flipflop7146/wizd.html
and downloaded the "Linux binary only: wizd.gz" and also the Full distribution.
I put the wizd executable in the root of the Flash, copied their wizd.conf file there also, and then added a line to the "go" script to invoke it. Took all of 5 minutes.
The program seems to be running and if I open a browser window and enter this:
//tower:8004
it does respond with a nice blank page. So it appears that it's running.
I don't know anything else about this server. It doesn't seem to take much memory, though perhaps that's because it's not really doing anything...
How can I determine if this is working correctly? Looking through the config file I see that you need to set up paths to your media - which is no problem. So if this is all there is to it, then it's very easy to run this on the UnRAID server.
limetech 10-18-05, 04:14 AM Hi Tom,
Feature Request:
Can we get access to the shares via WebDAV (in addition to SMB)?
Never heard of it. I went to their website & looked around - this will take more research so I put it on the "todo" list. Hint: more people asking for it moves it up the todo list :D
limetech 10-18-05, 04:23 AM This may have been asked earlier but I didn't see it, can two drives in the Unraid Server be put into a striped array (raid 0) so that the network will see them as a single drive? I would like to use 300Gb drives in the server, using 2 drives for music and the rest for video. I use a Slimdevices Squeezebox 2 for music playback and I can only use a single folder for playback via the squeezebox. This would require a single drive and I have more than 300 Gb of music burned so far.
Thanks,
Joe M.
If all the disks appeared as subdirectories under one share name, would that work with the Squeezebox? For example, suppose the share name was "Storage", then under My Network Places you would see:
Storage on Media Sever (Tower)
When you click on this, you would then see these directories:
disk1
disk2
disk3
etc.
So can you specify "Storage" as the directory, ie., "//tower/storage" and then would the Squeezebox find all your music in the "diskN" subdirectories?
limetech 10-18-05, 04:25 AM Here's a wacky thought for those who want to solve the shares problem....assuming that at the OS level you could have 1 master directory that each of the un-raid drives could be mounted under (with some sort of logical name) and THAT could be the share that is smb exported (samba?). You could still have a drive letter for EACH share, as well as a single unified drive that contains all of the drives are folders....eh, maybe that's how they do what he was describing
...
But rather than having multiple drives exported, you'd have 1 drive with these (or disk1, disk2, etc) as folders below that master share. Wouldn't require any fancy file system stuff other than mounting the drives systematically under a master folder and exporting that folder.
Rich
Hey that's a good idea :D
leonowski 10-18-05, 04:38 AM Never heard of it. I went to their website & looked around - this will take more research so I put it on the "todo" list. Hint: more people asking for it moves it up the todo list
Hi Tom,
Thanks for considering the request to add WebDAV support. WebDAV is a file sharing protocol (like SMB). In a nutshell, it allows you to read and write through HTTP. For example, WebDAV support for the Un-Raid server would be great - just point a browser or any WebDAV client to something like:
http://tower/disk1
Windows XP supports WebDAV. You can view WebDAV URLs as regular folders in Windows XP. More info on WebDAV is found here:
http://en.wikipedia.org/wiki/WebDAV
Thanks again!
Excellent continuing support Tom! Looks like WizD may be simple, dunno' what Twonky would be like or if it would even be advantageous. I'm willing to try out Wizd, I'm not married up to anything at this point. <shrug> Slicing off a reasonable portion of a drive for swap is a-okay with me! Thoughts posted regarding changes to the way shares are seen sound good, I've honestly not been using UNC naming - hrm...! I'm also not familier with WEBDAV but it sounds interesting as well.
I'm currently browsing around looking for HDs. I intend to use what I have on hand at first but would like to also purchase a solid drive for Parity. I know the Hitachi 500Gig drive is VERY attractive but it's pricepoint is a little high right now. From the sounds of it swapping out the Parity drive in the future is doable without too much fuss, I think I will wait patiently for prices to drop on the larger drives and then swap one in.
So, that leaves me looking at drives in the 300Gig range right now and I have some questions... Its there anything in particular that I ought to be looking at if the PCI bus is our bottleneck? For instance I'm looking at a Maxtor 300Gig drive with 16meg of onboard cache, is that cache likely to show an increase in throughput in your array? I'd asked earlier about caching controllers and I guess this falls along the same lines - is drive cache or controller cache an area where improvements might be seen?
I'm simply trying to find max bang for the buck here without being penny wise and pound foolish. If a few bux on onboard cache helps I'll spend it, otherwise I'll perhaps use that money to buy a 320Gig drive with smaller 8meg onboard cache. :)
Lastly, if the SATA code is to be released will that be using the onboard controllers? Would I be smarter to buy an SATA drive instead? Does the SATA interface relieve a bottleneck and increase performance out the GigE port or do we hit another wall? My understanding is that not all SATA drives have equal throughput either so buying one of those may take some research. <sigh>
Thanks!
P.S. Are you still only shipping out on Fridays (sob) or would an order placed say last night be shipped fairly quickly? I got lucky and most of my hardware is coming 2day :D
John Spicer 10-18-05, 11:20 AM Quite like to order one of these if you can sort out shipping to the UK.
boykster 10-18-05, 01:27 PM Hey that's a good idea :D
Do I get a cookie? :D
There is one caveat to this approach that I just remembered. Windows explorer will report free disk space incorrectly. Lets say that the primary mount folder is
/home/share/StorageServer
and resides on a disk with 10gigs of space. This is shared using samba as "storageserver" share. Each subsequent drive is mounted
/../../../disk1 (300gigs free)
/../../../disk2 (222gigs free)
/../../../disk3 (100gigs free)
..etc
When you connect to "storageserver" share it will report disk free of 10gigs (the mount folder's size) and when you navigate further down the tree into each subfolder, it does not update with the correct disk space :confused:
It does work, but for accurate disk space, you'd need to connect to a share that is located on each seperate disk. It works great for read-only clients however.
I don't know if it is a limitation on Samba and/or if it can be fixed, etc. I am running 3.0.8 and have never really looked for a solution.
FYI....
Rich
It does work, but for accurate disk space, you'd need to connect to a share that is located on each seperate disk. It works great for read-only clients however.
I don't know if it is a limitation on Samba and/or if it can be fixed, etc. I am running 3.0.8 and have never really looked for a solution.
FYI....
RichIn my opinion, since the only interface most will ever see is the amount of free space as reported to MS-Windows by Samba, this solution of a single share with subsequent drives mounted on subdirectories is not a very acceptable one if it does not accurately report on the space used/free.
Keep thinking of alternatives... we have a lot of collective brainpower and experience to draw upon. :confused:
Joe L.
boykster 10-18-05, 06:24 PM Unfortunately, so far I don't have a fix for the smb problem....I have tried substituting an alternative df directive, but that only seems to run on a per share basis, not as you walk the tree.
It is a known issue, but still have yet to find a resolution.
Rich
Okay, just fired up WizD under windows. Streamed video from a share on my Linkstation across my network to my wireless laptop and then back to my network to play on my Buffalo Linktheater. :lol: Yeah, not perfect performance but the interface isn't bad, plays MP3 pretty good so far. So yeah, I can probably live with this on my array if Twonky isn't of interest. I might also just run a Daemon on a box somewhere on the network accessing the array if I must - we can skin this cat many ways it seems :)
Following the discussoin about alternate ways to view the directory structure... I agree that viewing disk space is important. I'd consider giving that up in order to not have to view lots of shares though. If I could get accurate sizing by getting properties on the top level folder I'd be okay with that but it might not be super intuitive in say an office environment.
On a happy note - it seems my FOB has shipped and will be here Friday which is just in time for my hardware build! I still need info on HDs but I will have enough hardware here by then to at least get a start on things. Most cool, thanks!
limetech 10-19-05, 11:52 AM In my opinion, since the only interface most will ever see is the amount of free space as reported to MS-Windows by Samba, this solution of a single share with subsequent drives mounted on subdirectories is not a very acceptable one if it does not accurately report on the space used/free.
Keep thinking of alternatives... we have a lot of collective brainpower and experience to draw upon. :confused:
Joe L.
Viewing free space under Windows has always been tricky for me. For example, in Windows 2000, if you go to My Network Places, I don't see anywhere that free space is displayed, unless you first go to Computers Near Me, then select the Server, then select a share and you get a nice pie chart which shows Capacity, Used and Free for that share.
Under Windows XP I don't see any way at all to see Capacity/Used/Free for a share - does anyone know how to get this?
Regardless, the UnRAID management utility shows the Capacity and Free space for each data disk.
mapped drives show this from MyComputer, detail view.
boykster 10-19-05, 02:06 PM mapped drives show this from MyComputer, detail view.
On mapped drives you can also right-click the drive in My Computer to see the pie chart...total drive size, free space.
However, this only works if the share is mapped as a drive. If you use UNC mapping solely to access the share, then windows does not know anything re: the free space / capacity of the share.
After much further research (as a non-contributing samba developer, but a developer nonetheless) I have been unable to find a suitable fix for the tree-mount / free space issue I discussed above.
I do see this approach as a great way to consolodate N# of mounted drives into 1 share, especially for applications/systems that require a single point for media discovery. It also works great for "read-only" type applications/uses (ie client HTPC's reading dvd files / music files / HD transports / etc), but for administration / loading data I prefer using my seperate shares to get a good feel for free space. In my case (not using unRaid, but using linux/samba) this is the ONLY way I can get a feel for free space without logging into the linux box, as I am using 3ware RAID hardware/software which does a great job monitoring the hardware, but could car less about the filesystem on the drives.
Anyhow....good luck! I'll keep reading (already have tons of storage, even though I'm intrigued by this system) and if I find a solution independant of this thread to the mount/free space issue I'll let you guys know :cool:
Cheers,
Rich
curtis104 10-19-05, 04:58 PM Tom,
With my Media files on the UnRaid system will I still be able to stream my Media files through ORB Network.
limetech 10-19-05, 05:05 PM Tom,
With my Media files on the UnRaid system will I still be able to stream my Media files through ORB Network.
I don't know anything about ORB Network, but from their website FAQ I found this:
How does Orb find my media files?
Once Orb is installed on your home computer, it enables you to access the media files in your default "My Pictures, "My Music" and "My Videos" folders remotely. You can also easily configure Orb to access media located in any other local or network folders.
Anyhow....good luck! I'll keep reading (already have tons of storage, even though I'm intrigued by this system) and if I find a solution independant of this thread to the mount/free space issue I'll let you guys know :cool:
Cheers,
Rich
Watch out for Sun's ZFS. I'm waiting to see if Solaris 10 will support PCI-Express SATA controllers.
Imagine infinitely expandable storage all protected by 64bit checksums!
Instead of using 12 drive trays, could I simply use 4-in-3 adapters? Aside from being less convenient, would there be any other disadvantages to this approach? I guess my biggest concern would be with heat build up since each tray has its own fan vs. each 4-in-3 cage 'sharing' one large fan.
It looks like the trays are about 40% of the cost of a bare system... so I'm willing to sacrifice external access/convenience for cost savings, unless this would cause other problems.
Also, it appears that a 4-in-3 (x4) approach would allow for a total of 16 drives. Does the unRAID software allow for more than 12 drives? If so, would the recommended power supplies support the additional load? This configuration would bring the total system capacity to 7.5TB w/500GB drives. :cool:
Received my unRAID FOB today - quite fast shipping thanks! Unfortunatly it will likely be late today or likely tomorrow before my case, MB, and most other toys arrive for a check run. :(
The FOB is quite small so concerns about it sticking out and being bumped are lessened, I actually wouldn't mind having a few of these myself hanging around :) There are some setup instructions provided in hard copy that I'm not sure I've seen on the WEBsite but are pretty clear IMO, I don't know yet if these are also found on the drive.
Looking forward to setting this up :D
P.S. I wouldn't mind seeing how this might work on an SATA PCI Express setup myself as it might alleviate bottlenecks. Solaris 10 is not something I'm likely to want to run however.
limetech 10-20-05, 01:48 PM Watch out for Sun's ZFS. I'm waiting to see if Solaris 10 will support PCI-Express SATA controllers.
Imagine infinitely expandable storage all protected by 64bit checksums!
I used to work at Sun - we're not worried :D [just kidding!]
limetech 10-20-05, 01:56 PM Instead of using 12 drive trays, could I simply use 4-in-3 adapters? Aside from being less convenient, would there be any other disadvantages to this approach? I guess my biggest concern would be with heat build up since each tray has its own fan vs. each 4-in-3 cage 'sharing' one large fan.
Sure you can use those. We haven't done any thermal testing but it would probably be just fine. You don't have to put 4 drives in there; you can limit to 3 if you wanted. But the first time you need to replace a drive: :mad:
It looks like the trays are about 40% of the cost of a bare system... so I'm willing to sacrifice external access/convenience for cost savings, unless this would cause other problems.
Yes they are a large part of the cost.
Also, it appears that a 4-in-3 (x4) approach would allow for a total of 16 drives. Does the unRAID software allow for more than 12 drives? If so, would the recommended power supplies support the additional load? This configuration would bring the total system capacity to 7.5TB w/500GB drives. :cool:
The current software only supports 12 drives because the Promise bios only recognizes 2 cards. Maybe you could plug in a different type of card (e.g., SIIG), but we have never tried this.
Doing some research - there are SATA PCIe (http://www.addonics.com/products/host_controller/adsa3gpx1-e.asp) cards out there, some of them somewhat cheap. They don't usually support more than two (http://www.addonics.com/products/host_controller/adsa3gpx1-e.asp) ports it seems unless you get an expensive one intended for RAID. There are external SATA enclosures (http://www.macgurus.com/productpages/sata/satakits.php) that look interesting too. Of great interest are SATA "port multipliers (http://www.sata-io.org/portmultiplier.asp)". These take a single SATA port and multiply it by FOUR and if you use the right one you are supposed to get full bandwidth.
Current hardware for this array is pretty cheap (trays are a large part of it) and that's a HUGE feature. It would be interesting to see how much performance might be wrung from a PCIe SATA setup though since the PCI bus is apparently or bottleneck. I've not yet gotten my M/B but since it's Intel I'm betting there will be no chance to overclock the CPU or PCI bus to see if that helps - bummer. I suppose a good question might be - do we really need any added speed? It's quite possible that for the uses most of us will be interested in this array will be "good enough" on cheap(ish) hardware. There's still congestion at the Gig NIC as well which PCIe might not assist...
P.S. Microcenter is selling the Stacker case for $189 in silver. Considering the $40 in shipping I paid for my case on ZZF I'd have been bux ahead to have gotten one locally at Microcenter. Ah well, at least mine will be black and not silver :rolleyes:
erikatcuse 10-20-05, 02:40 PM Will there be a huge difference in performance between SATA and PATA? I might wait but with black friday just around the corner there should be some good HD deals.
The current software only supports 12 drives because the Promise bios only recognizes 2 cards. Maybe you could plug in a different type of card (e.g., SIIG), but we have never tried this.
I'm planning to wait for the SATA version of the unRAID software. Will the SATA cards have a similar limitation?
dstroot 10-21-05, 12:16 PM Any news on the SATA version? Web site still says 10/17 :)
Be careful with ZipZoomFly! I ordered parts Sunday, all but the case was listed as "free 2day shipping". It wasn't clear which 2day they wanted me to use so I chose UPS 2day. Today the 21st no parts despite my getting a shipping notice on the 17th. Looking up tracking status it seems they will be here the 24th! I check my order on ZZF and find that my shipping is listed as UPS GROUND. So much for 2day. I have no idea WTH happened here and I am NOT happy, I had hoped to be putting this together this weekend. I ordered items from EBAY and from Lime Technologies AFTER my ZZF order was placed and they were all here Thursday. GRR! :(
Thomas J. Coyle 10-21-05, 07:52 PM BLKMGK,
When I ordered from ZZF I just stayed with the generic 2nd day shipping without selecting a carrier. My order was placed on Thursday the 6th, shipped on Friday the 7th and arrived on Monday, the 10th. You can't ask for better than that for 2nd day shipping.
However, I suspect that because your case is probably being shipped by ground (UPS) because it is so heavy, they probably just packed the rest of your order with it instead of sending it by a 2nd day carrier for some reason.
Since ZZF is in Newark, CA and I live in LA, CA, it is not that far (350 miles) to ship an order to my doorstep. However, my most recent order came by Fedex air which must be used if it is to make it in two days. I think that if you had not selected UPS and had just stuck with the generic "2nd day", or had ordered the case separately since it is a ground shipment type of item, you would have had the rest of your order in two days. Just a thought.
Regards,
TCIII
Eh, I dunno' know how I ended up with week long shipping. If I had known there was a trick to this I'd have done it. Near as I could tell I was suposed to pick shipping and I tried to pick the one that wasn't so expensive. If I wasn't supposed to pick a shipping method they should have said so - what a PITA! There shouldn't be a trick to getting the advertised shipping rate....
reel_fan 10-22-05, 01:06 AM Does this system support JBOD, or do I have to have the parity disk?
I like living dangerously.
Tom (limetech),
First, I'd like to say congratulations on creating a novel (and successful!) solution to a niche problem. I just finished reading this entire thread and your willingness to share and explain how your product works is a welcome rarity in this forum.
However, I was a bit concerned with what I read a couple pages back when licensing and the GPL came up...
Ahhhh, the GPL..... You are correct that we wrote our own driver but it was done in two parts. One part uses derived code from the linux software raid package; the other part is proprietary. The first part we will eventually make public since it's a derived work. The two parts together make a single "module" which is not linked into the kernel, but is loaded at run time. The fact that the two parts themselves are "linked together" I guess is a grey area and I honestly don't know if that's "kosher" or not - if not we can certainly create two modules.
Please don't take violating the GPL lightly. You're distributing modified GPL software, but there's no mention on your website about when or where the source to those modifications will be available.
Based on comments you made early in the thread, I'm assuming the "un-Raid" driver is in fact a kernel module derived from the md and/or ide driver(s). However, your description of how your proprietary module and your non-proprietary module "link together" raises more concern. If your proprietary module depends on "hooks" into the kernel exposed by the non-proprietary module then you could potentially have a difficult time justifying a non-GPL license for either kernel module.
There's definitely grey area WRT proprietary kernel modules. I would suggest taking a look at http://www.atnf.csiro.au/people/rgooch/linux/docs/licensing.txt for an explanation Linus gave on the subject a few years back. More recent discussions of proprietary/non-proprietary mixed kernel modules can be found on the LKML with varying view points.
In the end I hope you don't take this the wrong way. I'm not suggesting you give away all your work for free. I'm also not suggesting that you can't have a proprietary kernel module, because its obvious you can. Just ask nvidia and ati (granted, their modules control proprietary hardware...) .
I think you've come up with a very cool product and I truly hope it continues to be a success.
-Brian
Geordon 10-22-05, 09:20 AM I have been following this thread for awhile, and really like this un-Raid method. However, I am hesitant about having data stored in a non-Microsoft disk format (i.e., not NTFS). Maybe I don't understand the on-disk structure, as I am not a Unix/Linux user. I would like to have the ability to add or remove drives from the array which can be connected to a Windows system.
I am looking at the un-Raid for use of digital image and video archiving -- not as a DVD/MP3 server, and don't want my data stored in a format dependent on a single device. For instance, if my un-Raid system goes kaput down the road and LimeTech isn't able to support it, how do I recover my data?
mpenton 10-22-05, 09:35 AM Going to the Limeware home page and reviewing the FAQ shows that data access to the individual drives is available even if the server fails (providing you have another computer).
I am looking at the un-Raid for use of digital image and video archiving -- not as a DVD/MP3 server, and don't want my data stored in a format dependent on a single device. For instance, if my un-Raid system goes kaput down the road and LimeTech isn't able to support it, how do I recover my data?Geordon,
Many of us share the same concern for our data... so does Tom at LimeTech.
It is interesting that you do not want to depend on a single device, interesting in that most normal hardware based RAID depend upon the specific RAID controller card, and the data on the drive is NOT a normal NTFS file system, but instead a NTFS file system that has been "striped" across your RAID array. (In other words, unreadable to most programs unless you use the one specific RAID controller card that wrote it)
LimeTech's unRaid allows you to read the data on almost any computer you can put the drives into as the file systems are standard. (just not Microsoft standard)
In his FAQ Tom gave the link to a driver that can be used under windows to read the ReiserFS disks. I would download it and keep it with your other recovery and backup disks for use in the worst case scenario. (it is a read-only driver, but you can get to your data from windows based machines)
The link Tom gave is http://p-nand-q.com/download/rfstool.html
As far as being able to read the disks under a different Linux OS, since the ReiserFS is supported under most versions of Linux, it would not be too hard to find a bootable CD image that could be booted from to allow normal access to these same disks without unRaid protection on any PC.
We might suggest that Tom consider an additional product in his product line. A bootable CD with a generic version of Linux that will mount and share on the lan the ReiserFS disks in a read-only mode. This CD would not include the unRaid software or loadable unRaid modules so it would not compete with his primary product. It would simply be a generic ReiserFS recovery aid and support generic disk controllers and SMB.
I did a quick google search... Here is a link to one such bootable recovery disk that seems to support the ReiserFS and samba to create network shares. http://www.sysresccd.org/ (I have not tried it, but it is only to serve as an example as it was the second link I found using Google)
I would love a bootable CD that can mount my ReiserFS drives (read-only) on a spare PC. It would mean that nobody will depend upon any single device to get to their unRaid archived data.
Any interest Tom, in adding such a "recovery disk" to your product line?
Joe L.
RADIatiON 10-22-05, 10:23 AM I have been following this thread for awhile, and really like this un-Raid method. However, I am hesitant about having data stored in a non-Microsoft disk format (i.e., not NTFS). Maybe I don't understand the on-disk structure, as I am not a Unix/Linux user. I would like to have the ability to add or remove drives from the array which can be connected to a Windows system.
I am looking at the un-Raid for use of digital image and video archiving -- not as a DVD/MP3 server, and don't want my data stored in a format dependent on a single device. For instance, if my un-Raid system goes kaput down the road and LimeTech isn't able to support it, how do I recover my data?
According to this earlier post (http://www.avsforum.com/avs-vb/showthread.php?p=6129569&&#post6129569) there is a simple driver you can install onto your windows systems that would allow you to read the data off any of these drives.
limetech... please correct me if I'm wrong ;)
...
A word about file systems. We originally wanted to use NTFS because we envisioned the possibility of moving disks between the server and individual windows machines. Though NTFS support in Linux is limited to read-only, there is a company that makes a "NTFS for Linux" commercial product. We abandoned this though because the liscensing was too expensive and besides it was dog slow.
Regarding ReiserFS: for those interested there is a lot of info out there you can google. There is also a free utility you can download that lets you read ReiserFS data under Windows. This is nice to have in case you need to get at some data on the Un-Raid and it's unusable for some reason.
Mark J. Foster 10-22-05, 12:16 PM Hi, Gang!
It's sure easy to find! Google "reiserfs windows", and the first hit is what you're looking for - it's called rfstool (http://p-nand-q.com/download/rfstool.html). If you're not comfortable with command line programs, you can pick up rfsgui (http://www.wolfsheep.com/map/) instead (scroll down the page until you reach it). Rfsgui is based on rfstool, but uses a more familiar graphical interface to let you read the disks.
Do note that both rfstool and rfsgui are read-only, so you can't write to reiserfs disks from Windows.
EDIT: Whoops! Sorry about that, Joe! I'd missed your post before responding.
Cheers!
MarkF
A bootable CD to read these disks is not needed from Tom, rather do a search on Knoppix (http://www.knoppix.org/) or if you're into INFOSEC look for Auditor based off of Knoppix. Knoppix will recognize most any hardware, doesn't modify your HDs unless asked, and will most certainly read ReiserFS without a problem. There are also apparently drivers for Windows that will allow it to read ReaiserFS which is one of the more robust filesystems out there. Considering what it would take to get NTFS working under Linux I'm MUCH more comfortable with this file system choice. There's also a newer ResierFS coming out that may be even better but it's not yet very well supported...
And yeah I do still have some concerns over the GPL licensing since it's not even mentioned anywhere on the site at all. I'm not a stickler for this but some are and this will become a problem if it's not addressed. If I were to begin advertising this in my .sig in some of the places I post I'm quite sure the heat would come :( This needs to be addressed sooner rather than later IMO and it needs to be clear.
The current software only supports 12 drives because the Promise bios only recognizes 2 cards. Maybe you could plug in a different type of card (e.g., SIIG), but we have never tried this.Tom,
Are you using the Promise bios on the IDE controller card? (is there even one on the controller card?), or are you referring to the support for the Promise IDE controllers built into the Linux kernel that limits support to two cards?
Joe L.
Mark Ward 10-22-05, 02:21 PM I'm also sold on this idea.
I currently have a Suse 9.1 Linux Server with 11 Drives (Boot drive + 10 Data Drives) in JBOD. I always figured that if any particular drive failed I have the original DVDs to reload from.
All my current drives are Reiserfs formatted (2 Sata & 9 IDE), and I have:-
Asus AV7N8X-E Deluxe Motherboard
Athlon 2500 CPU
1Gb Ram
Would I be able to incorperate my existing Drives within the array complete with their data? The idea of simply Flash booting my existing system sounds great. If I need to replace the Sata volumes it wouldn't be too disasterous.
Looks like a great product!
Mark.
Thomas J. Coyle 10-22-05, 02:27 PM Mark Ward,
I know that the Lime-Technology OS does support your Asus motherboard only the USB bootup is at 1.1 speed.
Since you already have your drives formatted in the proper format and loaded with data, you might only have to build the parity drive from the existing data drives.
Regards,
TCIII
Geordon 10-22-05, 05:57 PM Geordon,
Many of us share the same concern for our data... so does Tom at LimeTech.
It is interesting that you do not want to depend on a single device, interesting in that most normal hardware based RAID depend upon the specific RAID controller card, and the data on the drive is NOT a normal NTFS file system, but instead a NTFS file system that has been "striped" across your RAID array. (In other words, unreadable to most programs unless you use the one specific RAID controller card that wrote it) Joe L.
I did not say I currently have a RAID array -- for the reasons you list, I have not gone that route. As mentioned earlier in the threads, RAID is not a backup. However, I don't feel like backing up that amount of data, thus unRaid really interests me.
Thanks for all the responses to my post, pointing me in the right direction, and helping to alleviate my fears. Now, I just have to wait for the SATA support.
Mark Ward 10-22-05, 06:55 PM I know that the Lime-Technology OS does support your Asus motherboard only the USB bootup is at 1.1 speed.
Since you already have your drives formatted in the proper format and loaded with data, you might only have to build the parity drive from the existing data drives.This is what I was hoping to hear. This way I can simply remove my boot drive and replace it with a parity drive and I'm away! :)
I do however appreciate the performance advantage of the Intel motherboard, and that coolermaster case looks perfect. This way I can get the protection now and build a more efficient server at my leisure.
My current case Yeong Yang (http://www.yycase.com/yy-0221.htm) will hold a lot more drives (20 if using 5into3 Caddies), but that capacity just isn't necessary in this setup.
As well as a large part of my DVD collection my server also holds Acronis backups of all my other home machines, laptops and my office's business server. I'd be ok with the inconvenience for now, but I'd like to end up with no performance drop on DVD streaming whilst a 40gb+ Backup is writing to the Server from a different PC.
A couple of other questions...
1/ Is there any advantage using 2 smaller PSUs over 1 more powerful one?
2/ Is there really much of an electricity saving when the drives spin down? I'm aware my server currently uses a lot of electricity, but does less load on the PSU actually draw that much less power from the mains?
Thanks,
Mark.
CANNON-FODDER 10-22-05, 10:55 PM ---edit---
~6.4W (http://www.silentpcreview.com/article242-page4.html) in this chart.
Methodology (http://www.silentpcreview.com/article242-page3.html)
---/edit---
Here is one site's numbers.
Power Distribution within Six PC (http://www.silentpcreview.com/article265-page1.html)
SPCR's PSU Test Platform V.3 (http://www.silentpcreview.com/article269-page1.html)
Not vouching for the methodology, I hated EE back when...
Off Topic -- Because someone at work just bought 10 Shuttles with this GPU for [mostly 2d/text] workstations, my favorite part was:
The Aeolus 6800GT, on the other hand, drew much more power at idle — about five times as much. All of the power came from the +12V line; neither of the other two lines were affected. To put it in perspective, the entire A64-3200+ socket 754 system used less power at idle than this video card.
v/r,
C-F
limetech 10-24-05, 03:37 AM Will there be a huge difference in performance between SATA and PATA? I might wait but with black friday just around the corner there should be some good HD deals.
No "huge" difference in normal read/write operation. Everything else equal, SATA system does resync faster.
limetech 10-24-05, 03:57 AM I'm planning to wait for the SATA version of the unRAID software. Will the SATA cards have a similar limitation?
Current limitation of 12 drives is mainly imposed by Promise controllers. There are other SATA configs that can support up to 16 drives & that is the "new" limitation in the SATA version. Is that sufficiently vague? ;)
There are various ways to get to 16 drives:
- Use a motherboard with 8 SATA ports and a PCI controller with 8
- Use a motherboard with 8 ports and two 4-port PCI controllers
- Use a motherboard with 4 ports and three 4-port PCI controllers
- Use a motherboard with 4 ports, plus an 8-port and a 4-port PCI controller
You get the idea. Current, tested code supports 12 drives: motherboard with 4 ports and two 4-port PCI controllers. Other combinations listed above are possible depending on driver support and/or patches, but we have not tested any 16-port configs yet.
At some point it becomes valid to ask: just how many drives can practically comprise a single array/system? Most RAID systems limit array width to 8 or 12 drives (some fewer than 8), a very few 16. None that I know of support more than 16.
Here's the system/performance impact as more drives are added to the array:
a) takes longer to sync parity
b) takes longer to "reconstruct on-the-fly" data of a failed block (or entire disk)
If a) gets too long then you increase the "window of opportunity" for a second drive to fail while rebuilding a failed drive.
If b) gets too long, then you may not be able to provide a sufficient transfer rate to avoid "glitches" in a media stream when accessing data from a failed drive.
Personally, I'm confortable with 12 drives. I would consider 16; but I would not go over 16. Probably we will not offer any preconfigured products beyond 12 drives per array.
limetech 10-24-05, 04:02 AM Any news on the SATA version? Web site still says 10/17 :)
Yeah, I need to get around to updating the web site!
Status of SATA version is this: It's been running reliably for a couple weeks. But before releasing the product we have to make sure that a) it's truely reliable, and b) we're compliant with everything we need to be compliant with from a business perspective :) [Just trying to be honest with everyone.]
limetech 10-24-05, 04:05 AM Does this system support JBOD, or do I have to have the parity disk?
I like living dangerously.
If all you want is JBOD there are numerous linux distros out there that will let you easily build a JBOD. However, we will be adding an option to configure a system as a JBOD for those who want to use our software and packaging in a JBOD application.
limetech 10-24-05, 04:21 AM Tom (limetech),
First, I'd like to say congratulations on creating a novel (and successful!) solution to a niche problem. I just finished reading this entire thread and your willingness to share and explain how your product works is a welcome rarity in this forum.
However, I was a bit concerned with what I read a couple pages back when licensing and the GPL came up...
Please don't take violating the GPL lightly. You're distributing modified GPL software, but there's no mention on your website about when or where the source to those modifications will be available.
Based on comments you made early in the thread, I'm assuming the "un-Raid" driver is in fact a kernel module derived from the md and/or ide driver(s). However, your description of how your proprietary module and your non-proprietary module "link together" raises more concern. If your proprietary module depends on "hooks" into the kernel exposed by the non-proprietary module then you could potentially have a difficult time justifying a non-GPL license for either kernel module.
There's definitely grey area WRT proprietary kernel modules. I would suggest taking a look at http://www.atnf.csiro.au/people/rgooch/linux/docs/licensing.txt for an explanation Linus gave on the subject a few years back. More recent discussions of proprietary/non-proprietary mixed kernel modules can be found on the LKML with varying view points.
In the end I hope you don't take this the wrong way. I'm not suggesting you give away all your work for free. I'm also not suggesting that you can't have a proprietary kernel module, because its obvious you can. Just ask nvidia and ati (granted, their modules control proprietary hardware...) .
I think you've come up with a very cool product and I truly hope it continues to be a success.
-Brian
First, thank you for your reasoned comment on this subject. Let me say at the start that we fully appreciate the GPL. If it weren't for the linux platform, and the GPL under which source code is made available, our product and many other "embedded" products just wouldn't exist (e.g., TIVO comes to mind).
We are currently working with people knowledgable on GPL requirements, and I can assure everyone that we will do what is required.
limetech 10-24-05, 04:24 AM Tom,
Are you using the Promise bios on the IDE controller card? (is there even one on the controller card?), or are you referring to the support for the Promise IDE controllers built into the Linux kernel that limits support to two cards?
Joe L.
Both.
limetech 10-24-05, 04:35 AM I'm also sold on this idea.
I currently have a Suse 9.1 Linux Server with 11 Drives (Boot drive + 10 Data Drives) in JBOD. I always figured that if any particular drive failed I have the original DVDs to reload from.
All my current drives are Reiserfs formatted (2 Sata & 9 IDE), and I have:-
Asus AV7N8X-E Deluxe Motherboard
Athlon 2500 CPU
1Gb Ram
Would I be able to incorperate my existing Drives within the array complete with their data? The idea of simply Flash booting my existing system sounds great. If I need to replace the Sata volumes it wouldn't be too disasterous.
Looks like a great product!
Mark.
For the current shipping Flash, your preformatted file systems will be recognized only if they are in partition 1. If this is not the case with your drives, email us and maybe I can make a patch for you. Also, though we have run the s/w on that motherboard, it has not been thoroughly tested.
limetech 10-24-05, 04:41 AM This is what I was hoping to hear. This way I can simply remove my boot drive and replace it with a parity drive and I'm away! :)
I do however appreciate the performance advantage of the Intel motherboard, and that coolermaster case looks perfect. This way I can get the protection now and build a more efficient server at my leisure.
My current case Yeong Yang (http://www.yycase.com/yy-0221.htm) will hold a lot more drives (20 if using 5into3 Caddies), but that capacity just isn't necessary in this setup.
As well as a large part of my DVD collection my server also holds Acronis backups of all my other home machines, laptops and my office's business server. I'd be ok with the inconvenience for now, but I'd like to end up with no performance drop on DVD streaming whilst a 40gb+ Backup is writing to the Server from a different PC.
A couple of other questions...
1/ Is there any advantage using 2 smaller PSUs over 1 more powerful one?
Advantages:
1. Less expensive.
2. Able to turn one off if there are no installed drives connected to it.
Disadvantages:
1. Need a case that holds 2 supplies.
2. Cabling more elaborate.
2/ Is there really much of an electricity saving when the drives spin down? I'm aware my server currently uses a lot of electricity, but does less load on the PSU actually draw that much less power from the mains?
Good question. The main reason we spin the drives down is to save wear and tear on the drives.
albrigsr 10-24-05, 01:18 PM I have a little device called Kill-A-Watt that measure the Kilowatts per hour that a device draws. I will try to plug the server into it and have the drives spin down for several days versus not spinning down. My house is currently undergoing a remodel/addition so I can't promise I'll do his soon (or that I'll have electricity at all!) but that should help answer the question.
Heh, Slashdot (http://yro.slashdot.org/yro/05/10/24/1353218.shtml?tid=117&tid=17) has another article on a GPL violator up now. Man you do NOT want that kind of publicity. Makes for some good reading about the GPL though if you want a better nderstanding of some of the nuances IMO.
On a brighter note - my hardware *should* be on my doorstep when I arrive home tonight. I got a peek at the Coolermaster case this weekend and it looks to be pretty nice. I'm looking forward to putting this bad boy together! :) Looking around I didn't think I'd have too many drives to put in it but umm, I may quickly be up over 6 and needing that second P/S :eek: Going to be really nice to put these drives to use at last, I also purchased a 300Gig drive new for Parity. Will report back any issues I run into, I'm not expecting any at this point as the docs appear to be pretty complete. :cool:
erikatcuse 10-24-05, 02:02 PM Will there be a huge difference in performance between SATA and PATA? I might wait but with black friday just around the corner there should be some good HD deals.
Mark J. Foster 10-24-05, 02:12 PM Hi, Gotta5000,
Tom already answered your identical post above... :rolleyes:
No "huge" difference in normal read/write operation. Everything else equal, SATA system does resync faster.Best Regards,
MarkF
CANNON-FODDER 10-24-05, 02:22 PM Sorry for the poorly chosen links before, here is a better link to HHD power numbers, the previous pages were more about estimating the rest of the system.
~6.4W (http://www.silentpcreview.com/article242-page4.html) in this chart.
Methodology (http://www.silentpcreview.com/article242-page3.html)
Be interesting to see if the numbers there match with albrigsr's experiment.
v/r,
C-F
roscoe62 10-24-05, 06:31 PM It's great to hear you're still beavering away on the SATA version Tom. You continue to impress! :)
It's reached the stage now where I've got to ask a newbie question - but it relates more to Linux than to the Limetech solution.
If I'm using this system as my Media Server, do I need any other apps to act as a go between? What I mean is - when I've stored all of my backed up DVD files on the server, will they still be seen as the same filenames - eg VTS_01.IFO etc, or do I need a small app to act as an "interpreter"? If I rip my DVD files on my main PC and transfer them to the limetech server, will they just transfer across as is?
Not having any experience with Linux these questions only just occurred to me!
I don't mean to hijack the thread, but hopefully I'm not the only Linux newbie with these questions!
boykster 10-24-05, 07:44 PM It's great to hear you're still beavering away on the SATA version Tom. You continue to impress! :)
It's reached the stage now where I've got to ask a newbie question - but it relates more to Linux than to the Limetech solution.
If I'm using this system as my Media Server, do I need any other apps to act as a go between? What I mean is - when I've stored all of my backed up DVD files on the server, will they still be seen as the same filenames - eg VTS_01.IFO etc, or do I need a small app to act as an "interpreter"? If I rip my DVD files on my main PC and transfer them to the limetech server, will they just transfer across as is?
Not having any experience with Linux these questions only just occurred to me!
I don't mean to hijack the thread, but hopefully I'm not the only Linux newbie with these questions!
Not Tom, but familiar with linux/windows file sharing. A linux based file server, that makes shares available to windows networks, generally runs a server program called Samba. This emulates windows based SMB file sharing, and for all intents and purposes exposes disk shares that to your windows machine, look, act, feel, etc just like a share on a windows machine.
You will not need any middleware to access the files, and things should work just as if you shared a folder on another windows machine. Samba filters you from any major file system differences, and makes the environment seamless.
HTH,
Rich
~6.4Watts per drive when idle, but spinning.
CANNON-FODDER,
Multiply 6.4 Watts times 12 drives and you are looking at 76.8 watts of power you are not using when the drives are spun down. Now, this is not a lot of power, but it is roughly equivalent of leaving a 75 watt light bulb burning in your home 24 hours a day, 365 days a year. That equates to 657 kilowatt hours of usage.
My most recent electric bill here in North Carolina seems to indicate I pay roughly $6.75 per 100 kilowatt hours for power. That might indicate a potential saving of $45 per year if a fully loaded array is not spinning. That is enough savings over the course of a year to buy another three or four DVDs, or even another hard disk when they are on sale. (and somehow, I expect the electric rates to increase as fuel costs skyrocket, so my savings could even be more.)
Hey, it is not going to change my lifestyle, but $45 is still $45.
Joe L.
I could swear that someone posted a link to the D865GLCLK user manual and BIOS updates but damned if I can find it. Putting my box together and need to get some clarification on a few things. Google finds similiar but not exact boards - grr!
TIA!
Also - where's a good source for extended IDE cables? Looking at this thing I think the Cremax trays are also a really good idea for this monster! :eek:
limetech 10-24-05, 10:14 PM I could swear that someone posted a link to the D865GLCLK user manual and BIOS updates but damned if I can find it. Putting my box together and need to get some clarification on a few things. Google finds similiar but not exact boards - grr!
TIA!
Here you go: http://www.intel.com/products/motherboard/d865glc/index.htm
Click on the 'technical documents' tab; then click 'Product Specification'.
Also - where's a good source for extended IDE cables? Looking at this thing I think the Cremax trays are also a really good idea for this monster! :eek:
For a 24" cable: http://www.ewiz.com/detail.php?name=ST-ATA24
dougnliz 10-24-05, 10:21 PM dougnliz,
I assume that you ordered the Intel motherboard from zipzoomfly. I just received my motherboard from them today and unfortunately, you will find no documentation with the motherboard. However, there is a wealth of information about this motherboard on Intel's website: http://support.intel.com/products/motherboard/d865glc/index.htm. Hope this helps.
Regards,
TCIII
Here is the post with the link to the Intel motherboards.
I got my 24" cables from NewEgg. Fast shipping too.
Good luck!
Doug
roscoe62 10-24-05, 11:07 PM BOYKSTER,
Thanks for the answer. So can I safely assume the USB key donglie thingie (technical term ;) ) from Limetech will allow the server to operate in this fashion?
Thanks!
CANNON-FODDER 10-24-05, 11:25 PM CANNON-FODDER... Err, I was just answering the question, I do not have a dog in that fight.
I realized when the question was repeated that I did not link directly to a representative figure for each HHD itself.
v/r,
C-F
Thanks to both of you guys! I checked Mwave, Newegg, and eWIZ, all were pretty close - shipping adds a pile darn it. I may check local just to see if it makes sense to save the shipping. Pc-pitstop.com looks to be best for the Cremax parts so far. Putting this together I believe I have to agree that not using trays is insane but wow does it add to the cost if you go for good ones :eek: The LCD models are most attractive to me but it adds up, cannot help but wonder if the temp reading and alarms are really worth the added cost. Thoughts?
Sadly it looks like this box won't be getting fired up tonight. I'm still copying data off of drives and prepping things. Y-power cables and IDE cables may be needed along with some mounting hardware for lift off. 4 drives in this case is no biggie but to max out this box you either need trays or mounting kits - trays make LOTS of sense IMO so I'm shopping. On the plus side it turns out I had more drives of larger capacities than I'd realized. While this won't be "huge" I will likely have a good 6-8 drives in here for starters if I can rig a temporary mounting. Getting closer anyway!
Err, I was just answering the question, I do not have a dog in that fight.
No problem..., not fighting, just supporting my own savings.
Now, I only have 4 drives populated in my unRaid server, so my potential saving are somewhere near $1 per month.
Let's see... what did I pay the other day to get that cup of coffee at Starbucks???
Perhaps I can factor in the reduced cooling load on my air-conditioner... nah... better not tell anyone how much I spent in order to save $1 per month. :D
Guess I have to go along with Tom@LimeTech. Reduced wear and tear on the disk drives if they are not spinning... better reliability over time... data that is safe... yup, that's it.... that's why I have an unRaid server to begin with.
Joe L.
...Putting this together I believe I have to agree that not using trays is insane but wow does it add to the cost if you go for good ones...
Help me out – I might be missing something - why is it insane to not use the trays?
The lowest priced mobile racks on that site are $22. That is $264 for 12 plus shipping. The pros for mobile racks are it is easier to add and remove drives, the extra fans provide better cooling, and it looks good.
I plan on getting the CODEGEN 4U-500-CA Black 4U Rackmount Case that mpenton is using. I like rackmount better and it will cost less. I think case fans mounted next to the drives should keep them cool enough, but I guess I won’t know until I try it. I won’t be adding and removing drives that often so that doesn't justify the extra cost of the trays.
Edit: Please don't take any of this as negative - I'm just trying to avoid making any expensive mistakes :)
Help me out – I might be missing something - why is it insane to not use the trays?
I have to agree that 'insane' is a bit of an overstatement. ;) Call me a cheapskate, but I don't think cracking the case every now and then is such a big deal... especially to save ~$400 on 12 SATA trays.
Oy, I typed a response regarding trays at work and then left before posting - doh! :rolleyes:
So.. this case will hold 12 drives - there needs to be cooling. Out of the box it will mount 4 3.5 drives no issues, the rest of the slots are 5.25. Mounting drives there requires you to purchase mounts. Most of those mounts I've seen, unless you perhaps get the Coolermaster units, require you to screw the drive in and then you get to screw it into the case. Now, Coolermaster can mount these with just 2 screws but still - screws everywhere! Plus most of the generic mounts I've seen have zip for cooling, no shock mounting, and yeah there's those screws. Now you can buy units with cooling but those will cost more and may still have screws to take in and out. The drive trays from what I can tell mount a sleeve, have cooling, have an alarm to tell you if cooling fails, and you can get temp readouts and uptime as well. Mount 12 drives w\screws, run all those cables, and deal with it every time a drive changes plus worry cooling or... run trays. Yup, not cheap and yes I'm stressing over it. I will likely mount a few in trays and use the included and well ventilated cage for now. BTW - right now one drive is sort of free hanging in the case and is the same model as one in the cage. It's running 36 degrees vs the 26 for the identical drive in the cage. :eek:
Speaking of which - the array is now up and running, presently formated and synching at about 17megs a sec according to the WEB page. Setup was, overall, pretty easy. This appears to be using ReiserFS4 BTW which as I understand it is brand new and supposed to be pretty good. Most cool! Being able to name the shares might be nice, not delved deeper to see how that's done but I seem to recall it's possible. A single large share would still be awesome but this looks workable for now while testing etc. Time to start watching the sales papers for drives I guess :D
Hrm, and maybe getting WizD on the Flash and configured! :D
P.S. Trays I'm looking at are no more than $38 apiece for good ones and $22 or so for less good. The lower end of that scale isn't $400+
mpenton 10-25-05, 09:48 PM Right now I've got 11 hdds in my Codegen case with a 550 watt ps running the whole thing. The case supports a single 120 mm fan in front of the main array of drives (7 mounted vertically) with an 80 mm in front of a 2 to 3 adapter. The rear of the case has two 80 mm fans pulling air out. The highest drive temp I have observed is 34 degrees on the parity drive while transferring 400 gb of recorded hd material, the rest of the drives run from 22-28 degrees while in use. I used the Codegen case because of the best of reasons; I already owned a couple of them ;-) For those wondering about the ease of changing drives I'll help, it isn't easy, two screws for the case top, four for the drive cage (rubber isolated) and then the screws for the drive. The biggest part of a drive swap is connecting/dis-connecting the cables. I don't plan on running through drives so fast that it becomes a big issue. I will say that I have an intense dislike of the lower end drive trays because every one I've had experience with had a buzzy little fan that made way too much noise and eventually failed. YMMV
I've been warned by friends about the cheap drive cages already and Tom has also pointed the folly of using them. A friend of mine runs an array without parity or anything else and has heat issues - losing data in the process. I'm less worried about data loss than he is with parity in place but losing drives would suck. I'm currently using spares laying around to learn with but these WILL get upgraded as time and finances allow. Heat is my enemy for sure!
BTW - I cannot even hear this thing in my office right now. I run Shuttle XPCs so it's not roaring but they are apparently drowning out this thing. :eek: I'll be moving it to another room eventually so perhaps then I'll get a better gauge as to noise. For now it's NOT objectionable. 5 drives on board right now, case sides removed..
Richard_P_Harvey 10-26-05, 09:49 AM WOW - How in the world did I miss this thread when it got started.....! Anyway this is exactly what I have been looking to do for a long time now. I currently have 2 left over PC's acting as my media servers, using the available IDE channels in each and then adding external firewire cases and drive I have grown to now 12 total drives, some as small as 80GB and others much larger (320's). This would be a great solution to bring this all together and would allow me to migrate off the smaller drives as the get older and die OR great deals on big drives come along.
In reading this thread it's looking like most people are opting for more of the DIY path which I'm fine with. Can someone help me source this stuff at the best places.....? I too think the fully built price at Lime is high but if it's the olny way to get this thing all set up right I'm willing to do that too, saving a few $$$$ however would be good.
I do have one other question, all of my drives have data on them, mostly riped DVD's. If I build or buy this system will I be able to drop in these drives in and not loose any of the populated data that is on them...?
Mark J. Foster 10-26-05, 10:06 AM Hi, Robert!
Assuming that you're not running Linux and the Reiserfs file system on all those PCs, it will be necessary to move over gradually, since you'll have to copy the information off of each drive before installing it in the array. Start with two large blank drives: a data drive, and a parity drive. Copy all of the files off your next-largest drive onto those two, then you can install that drive in the array, where it will get reformatted with the Reiserfs file system. Continue that process until all the candidate drives are installed in the array, and you'll then have a much more manageable solution that's protected from single drive failures.
Cheers!
MarkF
In reading this thread it's looking like most people are opting for more of the DIY path which I'm fine with. Can someone help me source this stuff at the best places.....? I too think the fully built price at Lime is high but if it's the olny way to get this thing all set up right I'm willing to do that too, saving a few $$$$ however would be good.Richard,
Like most of the others following this thread I am not infinitely wealthy. (Have not seen Bill Gates chime in, besides, he might not opt for a Linux based solution anyway :D ) On the other hand, my free time is very valuable to me. I enjoy watching the DVDs I have in my collection and when an opportunity presents itself I will spend a few hours and immerse myself in a good movie.
I purchased an assembled server from Tom at LimeTechnology. I did it knowing he was getting paid for his time to assemble it, configure it, test it, package it for shipping, and to get it shipped to me. I even ordered it with a pair of the 500gig drives. (His price on the drives was almost exactly the same as the best price I found on pricewatch.com, so he was not marking them up much, if at all) He even installed the two drives and tested them in the server for 48hrs prior to shipping it to me. (There's 4 hours I saved in formatting and initializing those drives)
I felt his price for the assembled server was not out of line with the materials and effort involved. Yes, I could have sourced the pieces from various vendors, and yes, I have built most of the PCs in my house, so the assembly would not have been too difficult. But it probably would have taken me the better part of a day to put it all together after I spent an hour or two researching and ordering the various piece parts needed.
Now, if I had more free time, or elected to spend my free time assembling the server rather than with my family and friends, or if the difference in cost was a lot greater, then the DIY route would have been my choice too. It is all about choices and Tom has given us those choices.
We can spend our time, or for a bit more money, his time. For me, all I had to do was unpack it and plug it in. I did add several more drives already, a task made very simple because of the drive trays, and will migrate more as I find the time to take apart the other PCs in my home and relocate the drives. For me, the drive trays will make this simple.
In the meantime, my unRaid server has been on my LAN for about a week now, its disks spending most of the time "spun-down." This past week my wife and I have watched several movies, taken some dance lessons, worked on making Halloween costumes, and enjoyed more of our very valuable free time doing things together. That was worth spending a few $$$$ more to me, even though I love to save money too and have done many DIY projects for my home theater. (See my prior post on how I now save nearly $1 per month since my drives go to sleep when idle :D)
Joe L.
erikatcuse 10-26-05, 12:57 PM Ok I read through the entire thread and found out all the hardware people are using for the server setup...I'll be building my own in the next few months and wanted to see what everyone was using. An * means this is what I believe limetech is shipping with.
Cases Used
* Coolermaster Stacker case STC-T01
CODEGEN 4U-500-CA (Rackmount Case)
PSU
*2x Sparkle/Fortron FSP300-60PN
Antec 350watt Smart Power Supplies
Drive Bays
*Cremax MB123AK
Kingwin KF-101
2-to-3 adapter (For Rackmounted Case)
4 into 3 adapter
Controller Card
*Promise Ultra 100 TX2
*Promise Ultra 133 TX2 (No advantage over the 100)
Maxtor ATA133 (Works but is really slow could flash to generic bios not tested)
Mother Board
*Intel D865GLCLK (or the 6 pci-slot variant) (LK = PRO/1000 GigEthernet chip)
Asus A7N8X-E Deluxe (AMD works ok not tested fully yet)
CPU
*Intel 2.26GHz Celeron D 315 Prescott 533MHz FSB Socket 478 Processor
Memory
*512 (256x2 DDR266) Min and Recommended
1gig (512x2) Would be fine no real increase dut to PCI bus limitations
Max upto mobo support (Not needed)
Hard Drives
Any type of PATA drive you want...the largest being your parity drive
IDE Cables For coolmaster stacker
4 18"
2 24"
albrigsr 10-26-05, 02:20 PM Hi, Robert!
Assuming that you're not running Linux and the Reiserfs file system on all those PCs, it will be necessary to move over gradually, since you'll have to copy the information off of each drive before installing it in the array. Start with two large blank drives: a data drive, and a parity drive. Copy all of the files off your next-largest drive onto those two, then you can install that drive in the array, where it will get reformatted with the Reiserfs file system. Continue that process until all the candidate drives are installed in the array, and you'll then have a much more manageable solution that's protected from single drive failures.
Cheers!
MarkF
This is what I had to do and a HUGE reason why I am a BIG fan of the removable drive bays!
erikatcuse 10-26-05, 02:27 PM Does the unraid server support an UPS so if the power goes out the server shuts down? Also how does the server report a drive failure. I'm going to be keeping the server in my storage closet and it won't be hooked upto a monitor after initial install
Does the unraid server support an UPS so if the power goes out the server shuts down? Also how does the server report a drive failure. I'm going to be keeping the server in my storage closet and it won't be hooked upto a monitor after initial install
There was mention of UPS support a few pages back but so far I've seen no documentation of that - I too am interested as it seems I've just scored a monster APC UPS for my box :D
Drive failures are reported on the WEB status page. There's no sense in having a monitor hooked up to these boxes as there's little on the screen shown except shell commands issued from the WEB frontend. I'll likely be removing my keyboard and monitor here pretty soon on mine. It would be nice if some sort of YAC (http://sunflowerhead.com/software/yac/) program works out for drive failure notification - this has been discussed some already.
Tom, any thoughts? Can we have a little documentation on how best to add SATA drives into the array too? Now that mine is running I'm pondering trying it out - maybe just for the Parity drive at least....
Thomas J. Coyle 10-26-05, 03:29 PM gotta5000,
Under "Drive bays" that should be 4 into 3 for the CM Stacker case.
I kept the front control panel on the CM Stacker by using eight Kingwin KF-101 mobile racks and one of the 4 into 3 drive subchassis that comes with the Stacker.
I have 250 to 300GB drives in the mobile racks and four 200GB drives in the 4 into 3 subchassis.
ZZF presently has the Intel motherboards for $94.99 with free 2nd day shipping. This is cheaper than any of the sellers on eBay though I managed to score one for $87 with free shipping, but the seller has sold out.
Regards,
TCIII
It would be nice if some sort of YAC (http://sunflowerhead.com/software/yac/) program works out for drive failure notification - this has been discussed some already.
I added a small shell script to my unRaid server to do exactly this. It is run hourly by cron (the linux cronological scheduler) and will report to YAK clients via a TCP/IP message across my LAN if a failure occurs.
I'm at work now, but I'll clean the script up (make it self-installing) and post how I did it when I get home later this evening if you are interested.
I shared what I did with Tom at LimeTech and his comment was "Neat script!" It is easy if you are experienced in the Unix environment, but slightly complicated in that the "filesystem" in the unRaid server is re-created on the fly every time the server is re-booted. (everything is in ram) Therefore, code needs to be added to the flash drive that "installs" the script in the cron.hourly directory every time the server is re-booted.
I'll post details tonight.
The ability for experienced unix users to add/customize the server is a major advantage of the unRaid server over some of the other NAS storage competition. Although the script is very simple, it adds a lot to my comfort level knowing that any disk failure will be brought to my attention in a reasonable timeframe.
Joe L.
Thomas J. Coyle 10-26-05, 04:07 PM J. L.,
Sounds like a nice addition to an already great OS! Did Tom say if he was going to integrate it into his next release of the UnRaid OS software?
Regards,
TCIII
erikatcuse 10-26-05, 04:34 PM Would you buy referbished? I've found the mobo for $65 and ultra133 for $22 at chiefvalue.com with $1 shipping
http://www.chiefvalue.com/app/productlist.asp?linkid=100&submit=search&description=Intel%20D865GLCLK
http://www.chiefvalue.com/app/productdetails.asp?submit=search&item=16-102-007R
Doesn't look like the mobo is in stock
Thomas J. Coyle 10-26-05, 05:01 PM gotta5000,
Depends on the refurb warranty period (90 days?) versus a new (OEM) motherboard warranty of 12 months and a retail unit warranty of 3 years.
I was able to buy two new Promise ATA 100/TX2 controllers on eBay for less than $20 each including shipping. There seems to be number of new ones for sale on eBay and there is very little competitive bidding on these units.
Regards,
TCIII
Joe, sounds like an excellent hack and I'm VERY interested in seeing it please! If nothing understanding how stuff is added and executed will be helpful.
I've managed to get WizD running on my box however it's not yet executing on it's own - I think I can get that going though, for now I execute it from the console (yuck). Where I'm having the most trouble is setting up aliases in wizd.conf. Normally you'd give it a drive letter and path but umm, not so here obviously. UNC naming didn't work either and I don't want to try a hardcoded IP if I can help it. It may require pathing to the Linux mounted disk but I'm not yet familier with where those are mounted and am *not* terribly Linux savvy so any insight provided by anyone else would be nice. If I make any decent progress here I'll let folks know. For now I have the wizd executable, conf file, docroot folder, and skin folder loaded on the FOB. I left off the HD skin though. This is working well enough to get me streamed audio through a shortcut to my Buffalo LinkTheater but nothing off the server itself. Grr! :)
P.S. My Promise controllers were as low as $13 shipped off of eBAY :D
Thomas J. Coyle 10-26-05, 07:47 PM Tom,
How is the second release of the UnRaid OS software coming? Do you have a projected release date yet?
Regards,
TCIII
Thomas J. Coyle 10-26-05, 09:42 PM Hi all,
If you are going to use the Intel motherboard, do not forget to check the BIOS version. It should be P24 as Lime Technology has indicated on their website. If it isn't, then you need to update the BIOS.
My first Intel motherboard was a series 409 and had the P24 version of the BIOS. However, my second motherboard, which is a newer series 413, had a P21 version of the BIOS. Go figure!
So I went to the Intel webpage for the D865GLCLK and down loaded their bootable BIOS utility and P24 BIOS image. This method uses a bootable floppy disk and requires that a floppy drive be hooked to the motherboard. Since my motherboard test fixture has a floppy drive on it, I took the bootable disk route.
From the Intel motherboard BIOS update download, I made a bootable floppy disk and followed the instructions on the BIOS update webpage. Everything went like a charm and the motherboard BIOS is now version P24.
Regards,
TCIII
Joe, sounds like an excellent hack and I'm VERY interested in seeing it please! If nothing understanding how stuff is added and executed will be helpful.BLKMGK & TCIII,
Here is the script I wrote to send YAC alerts to PCs on my LAN from my unRaid server.
Before you do anything... you will need to edit an existing file on the unRaid server. It is possible for you to mess up the existing contents and make the unRaid server unbootable. If you are not comfortable editing files and adding lines, do not try this. You have been warned... I am not responsible if you make your flash drive unusable.
With the warnings out of the way, this is actually not too hard and as long as you can add a line to an existing text file using notepad without changing other lines in the same file you will be OK.
Before you do anything on the unRaid server, download and install a YAC listener on your PC. You can get it from http://sunflowerhead.com/software/yac/
An alternative YAC listener is available as "Whoisit.zip" available here : http://forums.snapstream.com/vb/showthread.php?p=137721 It can be used if you are running WinXP or have installed the .NET runtime package as listed in the thread describing Whoisit.
I like the Whoisit appearance better than stock YAC, but it seems to cut the last character off the sent message (you will not see the trailing period of the error messages) Hey.. if it bothers you, use the stock "YAC listener" Add a shortcut to whatever YAC listener you use to your "startup" folder on your PCs. (you cannot run both on the same PC at the same time)
To create the unRaid alert in the case of a disk failure I had to find out the status of the unRaid disks. I found that the /proc/mdcmd "file" contained the current status. (The /proc folder does not contain real files, but system information in the form of files to make it easier to access) To see all the content type:
cat /proc/mdcmd
Once I found where the status was available I had to figure out how to send a TCP/IP message from the unRaid server to the YAC Clients running on my PCs. To send the message I use a copy of "netcat"
It is a general purpose unix utility that can send to (or listen to) any TCP or UDP port.
Now, netcat is not supplied by Tom, at least not yet (I've asked him to add it) but you can download a copy from this link (http://www.metesek.com/projects/trusted-binaries/files/binaries/netcat/bin/netcat)
Copy the "netcat" program to your /boot share drive on the unRaid server.
Next, download and unzip the check_unraid.sh script I've attached to this post.
Open the check_unraid.sh script in notepad and edit the second line to change the PC names to those on your LAN. If you only have one PC, then it can be the only name listed. If you have more than one machine separate the names with spaces. Note that you do not specify IP address, but machine names. The actual IP address will be dynamically looked up as needed.
Once you have edited the machine list put a copy of the script in the /boot share drive on the unRaid server.
now... Log onto the unRaid server using telnet and type the following unix commands:
cd /boot
chmod +x check_unraid.sh
chmod +x netcat
At this point all that is left is to copy the check_unraid.sh script to a folder where it will be executed hourly by the linux cronological scheduler (cron) To do this type the following command:
cp /boot/check_unraid.sh /etc/cron.hourly/
now, make a backup copy of the "go" script using the following command:
cp /boot/go /boot/go.original
One last step is to add that same "cp /boot/check_unraid.sh /etc/cron.hourly/" line to the end of the "go" script in the /boot folder. This will automatically install (copy) the check_unraid.sh script to the /etc/cron.hourly folder every time the unRaid server is restarted.
You can use notepad in windows to do this. The line shown below in red is the line you will be adding to the "go" script.
The last few lines in the "go" script will then be:
# Start the management utility
/usr/local/sbin/emhttp
cp /boot/check_unraid.sh /etc/cron.hourly/
Do not change any of the other lines in the "go" script... you have been warned.
I'll attempt to describe its operation.
Basically, the "grep" and "egrep" programs scan a file looking for text matching a given pattern. They set a status variable ($?) to 0 if the pattern was matched and to 1 if not matched.
I first scan /proc/mdcmd for the mdState=STARTED line. If it is not found I set an error message stating that the unRaid array is not Started.
I then scan for either =DISK_INVALID or =DISK_DSBL
If either is present I set an error message that the unRaid array needs attention.
Lastly, if I had set an error message, for each machine name in the notify list I use the net lookup command to find the IP address to use with the netcat command. It is invoked with the command arguments to exit once it has finished sending the message and to only wait 1 second to make any given connection.
If this is over your level of understanding get a unix/linux friend to assist you. Do not call Tom at LimeTechnology. If you make your server unbootable, then call him, but you better be real polite to him as you plead for assistance :D Odds are he can always sell you another flash-drive with a new copy of unRaid :D
Joe L.
# Define the machines you wish to broadcast an alert to here
notify_machines="htpc htpc2 jalp3 dellcpx c800"
yac_port=10629
PATH=$PATH:/boot/
# initialize the error message to an empty string
emsg=""
grep "mdState=STARTED" /proc/mdcmd >/dev/null 2>&1
if [ $? != 0 ]
then
emsg="unRaid array not started."
fi
egrep "=DISK_INVALID|=DISK_DSBL" /proc/mdcmd >/dev/null 2>&1
if [ $? = 0 ]
then
emsg="The unRaid array needs attention. One or more disks are disabled."
fi
# if an error message was set, broadcast it to all the machines
# in the notify list in turn
if [ "$emsg" != "" ]
then
# notify each machine on the notify list in turn
for i in $notify_machines
do
# look up the ip address given the machine name
ip_addr=`net lookup $i 2>/dev/null`
if [ $? = 0 ]
then
echo "$emsg" | netcat -w 1 -c $ip_addr $yac_port
fi
done
fi
Heh, Slashdot (http://yro.slashdot.org/yro/05/10/24/1353218.shtml?tid=117&tid=17) has another article on a GPL violator up now. Man you do NOT want that kind of publicity.
Exactly.
I considered submitting this to slashdot, but with the lack of details on the lime-technology site and no mention at all about the GPL or licensing of their software, they'd take quite a beating. It can be hard to recover from the reputation created by that kind of publicity.
limetech 10-27-05, 04:42 AM I have to agree that 'insane' is a bit of an overstatement. ;) Call me a cheapskate, but I don't think cracking the case every now and then is such a big deal... especially to save ~$400 on 12 SATA trays.
Definately if you're comfortable building your own system you probably won't mind the occasional aggravation of installing and/or replacing hard drives bolted into your case :)
BTW the Cremax mobile racks can be had for under $20/each.
As an aside - our very first idea for a case was to modify an Antec Sonata to fit more drives. We even tried to talk to Antec to see if they would do it :p Anyone who has that case would know what I mean - they have a great way of mounting the hard drives in there.
limetech 10-27-05, 04:46 AM Not Tom, but familiar with linux/windows file sharing. A linux based file server, that makes shares available to windows networks, generally runs a server program called Samba. This emulates windows based SMB file sharing, and for all intents and purposes exposes disk shares that to your windows machine, look, act, feel, etc just like a share on a windows machine.
You will not need any middleware to access the files, and things should work just as if you shared a folder on another windows machine. Samba filters you from any major file system differences, and makes the environment seamless.
HTH,
Rich
Exactly right. I would add that SMB/CIFS is also available on Mac's.
This appears to be using ReiserFS4 BTW which as I understand it is brand new and supposed to be pretty good.
From the first page of the thread:
We're using the 2.4.29 kernel...
I really hope that one of those two statements (or both) are incorrect. Lime Tech might be crazy enough to violate the GPL, but I don't think they'd be crazy enough to attempt a backport of reiser4 to the 2.4 kernel :eek:
Heck, I still wouldn't be totally comfortable with reiser4 on a 2.6 kernel. It's still very much in development...
Actually, now that I think about it, does the parity drive have a file system on it as well? I would guess not, since from info gathered here (if only www.lime-technology.com had half the info this thread does ;) ) the parity drive can't be any arbitrary drive, ie. it has to be the first drive in the array (/dev/hda). And since the XOR is done in kernel space, it would be much more efficient to do I/O at the block level...
Hmm... I can't wait to see the source to this "driver" so I can stop trying to think of ways to do it myself. :o
limetech 10-27-05, 04:49 AM BOYKSTER,
Thanks for the answer. So can I safely assume the USB key donglie thingie (technical term ;) ) from Limetech will allow the server to operate in this fashion?
Thanks!
Yes that's right. Our goal is to hide linux as much as possible (while Joe's goal is to teach you bash shell programming :cool: )
limetech 10-27-05, 05:00 AM Oy, I typed a response regarding trays at work and then left before posting - doh! :rolleyes:
So.. this case will hold 12 drives - there needs to be cooling. Out of the box it will mount 4 3.5 drives no issues, the rest of the slots are 5.25. Mounting drives there requires you to purchase mounts. Most of those mounts I've seen, unless you perhaps get the Coolermaster units, require you to screw the drive in and then you get to screw it into the case. Now, Coolermaster can mount these with just 2 screws but still - screws everywhere! Plus most of the generic mounts I've seen have zip for cooling, no shock mounting, and yeah there's those screws. Now you can buy units with cooling but those will cost more and may still have screws to take in and out. The drive trays from what I can tell mount a sleeve, have cooling, have an alarm to tell you if cooling fails, and you can get temp readouts and uptime as well. Mount 12 drives w\screws, run all those cables, and deal with it every time a drive changes plus worry cooling or... run trays. Yup, not cheap and yes I'm stressing over it. I will likely mount a few in trays and use the included and well ventilated cage for now. BTW - right now one drive is sort of free hanging in the case and is the same model as one in the cage. It's running 36 degrees vs the 26 for the identical drive in the cage. :eek:
You can also get those really cheap rails that you attach to the hard drive to enable you to mount it a 5.25" bay.
Speaking of which - the array is now up and running, presently formated and synching at about 17megs a sec according to the WEB page. Setup was, overall, pretty easy. This appears to be using ReiserFS4 BTW which as I understand it is brand new and supposed to be pretty good. Most cool!
No - it's definately ReiserFS 3.6
Being able to name the shares might be nice, not delved deeper to see how that's done but I seem to recall it's possible.
Not in the currently shipping s/w.
A single large share would still be awesome but this looks workable for now while testing etc. Time to start watching the sales papers for drives I guess :D
Hrm, and maybe getting WizD on the Flash and configured! :D
P.S. Trays I'm looking at are no more than $38 apiece for good ones and $22 or so for less good. The lower end of that scale isn't $400+
Be sure and post results of using WizD. If you run into problems email me and I'll try and help.
limetech 10-27-05, 05:11 AM Ok I read through the entire thread and found out all the hardware people are using for the server setup...I'll be building my own in the next few months and wanted to see what everyone was using. An * means this is what I believe limetech is shipping with.
Cases Used
* Coolermaster Stacker case STC-T01
CODEGEN 4U-500-CA (Rackmount Case)
PSU
*2x Sparkle/Fortron FSP300-60PN
Antec 350watt Smart Power Supplies
Drive Bays
*Cremax MB123AK
Kingwin KF-101
2-to-3 adapter (For Rackmounted Case)
4 into 3 adapter
Controller Card
*Promise Ultra 100 TX2
*Promise Ultra 133 TX2
Maxtor ATA133 (Works but is really slow could flash to generic bios not tested)
Mother Board
*Intel D865GLCLK (or the 6 pci-slot variant) (LK = PRO/1000 GigEthernet chip)
Asus A7N8X-E Deluxe (AMD works ok not tested fully yet)
CPU
*Intel 2.26GHz Celeron D 315 Prescott 533MHz FSB Socket 478 Processor
Memory
*512 (256x2 DDR266) Min and Recommended
1gig (512x2) Would be fine no real increase dut to PCI bus limitations
Max upto mobo support (Not needed)
Hard Drives
Any type of PATA drive you want...the largest being your parity drive
IDE Cables For coolmaster stacker
4 18"
2 24"
Hey you're pretty close there! I would add this:
* A couple power Y-Splitters unless you find 2 p/s with 6 connectors each
* A pile of cable ties to keep things neat
* A 6-pack of your beverage of choice to consume during assembly :)
limetech 10-27-05, 05:22 AM There was mention of UPS support a few pages back but so far I've seen no documentation of that - I too am interested as it seems I've just scored a monster APC UPS for my box :D
What's the model number of your UPS?
Drive failures are reported on the WEB status page. There's no sense in having a monitor hooked up to these boxes as there's little on the screen shown except shell commands issued from the WEB frontend. I'll likely be removing my keyboard and monitor here pretty soon on mine. It would be nice if some sort of YAC (http://sunflowerhead.com/software/yac/) program works out for drive failure notification - this has been discussed some already.
Right, no real need to have a monitor and keyboard hooked up.
Tom, any thoughts? Can we have a little documentation on how best to add SATA drives into the array too? Now that mine is running I'm pondering trying it out - maybe just for the Parity drive at least....
A caution: the s/w you have does not support SATA. If you look at the bios setup you will see that the Drive Configuration is set to [Legacy]. This setting disables the SATA ports. If you change this setting the s/w will not operate correctly.
Joe's goal is to teach you bash shell programming :cool: Tom,
I debated including the description of the shell script I wrote and decided it might save me from having to answer a few PMs from unexperienced owners attempting to add this feature. I'm not sure I can teach "bash" shell programming in a single post in this thread :eek: although it is very similar to the Steve Bourne (sh) and David Korn (ksh) shells I've used under Unix for the past 25 years or so.
Now that I've figured out a simple way to include alert-messages on the unRaid server perhaps you can include something similar and add a field on your setup page to enter the list of machine names to notify. (you would need to add "netcat" to one of the "bin" directories, but the rest probably wouldn't take too much time)
Clearly, I'm not the only person who wants to be notified in the event of a disk failure instead of constantly having to go look to see if everything is OK.
By the way... if you want an hourly notification that all is OK, change the the two lines at the top of the script from:
# initialize the error message to an empty string
emsg=""
to
# initialize the error message to an OK status message
emsg="unRaid is OK."
You will need to change the script in the /boot directory and then copy it again to the /etc/cron.hourly directory by typing the following command after editing the script to change the emsg initialization line:
cp /boot/check_unraid.sh /etc/cron.hourly/
I've done this on mine just to keep an eye on the status for now. I'll change it back when it becomes annoying that it keeps telling me everything is OK. :eek:
One last thing, the hourly notification occurs at 47 minutes past the hour. That appears to be the way the current "cron" task is defined and the scripts in the /etc/cron.hourly directory executed.
Joe L.
Exactly.
I considered submitting this to slashdot, but with the lack of details on the lime-technology site and no mention at all about the GPL or licensing of their software, they'd take quite a beating. It can be hard to recover from the reputation created by that kind of publicity.
Yes exactly! I had been about to add an URL to my .sig in a few places like Slashdot but man it would only take one zealot to really ruin things! :(
Bdogg - I wouldn't count on seeing the source of their driver. If they have done it correctly that portion will be distributed binary only and closed source with no static linking blah blah which shoudl make it GPL compatible. I've obviously got the software now but it's in a BZ compressed form and umm being a Linux noob I've not been sure how to pull it apart to add things to the image or examine what's going on. <shrug> It does work quite well though! Looking at the mounts I've not been able to figure out how to get to the mounted drives via Telnet either - so frustrating! Yes, noob, but dang it the Mount command isn't showing me anything sensible. Eh, no worries as it's functioning well but getting this info might help me with WizD... :p
My mistake on ResierFS4. When the machine was booting up and things were flipping by I could swear I'd seen a mention of it. Having done some reading on it I was actually pretty excited that this might be the case although yes I too will admit that the 4th release is still a bit new to be using on something critical. :eek: anyway, sorry for the misinformation concerning the filesystem.
Joe - thank you for both the files and the tutorial! So far I've yet to have luck booting and getting WizD to launch so if I can get your code launching then that will help. From what I can tell the /boot directory is a copy of the image off of the USB FOB - yes? I had added a line to the GO script but no joy yet, it may be a pathing problem perhaps. cp is a copy command yes? Looks like you are adding a command to a CRON job there, will be the first time I've ever done that (yes I'm a Linux noob). Very cool! As for making the FOB a brick - no worries. I backed mine up as soon as I got it. Since Tom will eventually distribute updates that will be as easy to add as copying files over the structure etc. of the FOB doesn't seem too critical plus it's FAT. So long as backups are retained and we don't crash the OS I think monkeying around isn't toooo dangerous :D
Tom - WizD is launchable from the console just fine. I think I can get it launching on it's own tonight after I get Joe's code running. However that's not really the big problem - the big problem I'm running into is getting a link to my media collection. The .conf file would normally list something like c:\media but obviously that's windows. It's Linux capable too but I'm not sure how to link to one of my disks with media on it in the context that WizD is running in. I am guessing that I need the mount point and pathing, so far with my level of knowledge that's been a no-go. I tried UNC mapping, that was a nog-go but I expected that. I *am* able to peruse my HTML file of links correctly at least but no streaming of media just yet. If I could get some help there I'd be rolling I think and able to see how well the system handles it. Nothing critical on my array yet so if it tries to die I can recover ;)
P.S. Gotcha' on the SATA support - no worries. I look forward to trying it out someday. Was pretty neat to arrive home last night to find all drives spun down and idling for a little savings :)
Looking at the mounts I've not been able to figure out how to get to the mounted drives via Telnet either - so frustrating! Yes, noob, but dang it the Mount command isn't showing me anything sensible. BLKMGK,
I'm pretty sure the USB drive is mounted on /boot so files added there are actually on the flash drive.
The other (data) disk drives are accessible via telnet in the /mnt/disk1, /mnt/disk2, /mnt/disk3, etc. directories.
I don't know why they do not show up when using the mount command either... but at least you can easily get to them under linux. When initially doing my shell script development I was doing it via telnet in the /mnt/disk1 directory. Once I had something I liked, I moved it to the flash drive on /boot.
I'm not adding a task to cron. There was already an hourly cron job that executed each script in turn that exists in the /etc/cron.hourly directory. By putting a copy of the unraid alert script there, it is executed on that same schedule. (that existing cron job runs at 47 minutes past the hour)
Programmatically adding a new task to the cron scheduler is not hard, but way harder than adding a copy of a script to a directory where scripts are executed already. I'll let you learn about adding a new task to cron on your own for now... be very careful, it is very easy to wipe out the entire list of scheduled tasks. (although it would be replaced as it originally was when you reboot from the flash drive, so your potential for problems are not as great as on standard unix/linux systems.)
Joe L.
Thank you for the clarification Joe! That helps and makes sense - I'll forge ahead on this tonight and let folks know how it goes. I suppose since I'm one of the ones bitching loudest for a media applet it only makes sense that I try to solve the problem first :rolleyes: Appreciate the info on CRON too, I'm fine with an hourly task on that schedule. Seems I now need to install clients and get that new modem hooked up somewhere so I can get CallerID too. I *really* want that on my TIVO too, yes it's already hacked up so not many worries there. Fun Fun!
curtis104 10-27-05, 10:19 AM Could you tell me what the downside would be using the Asus A7N8X-E Deluxe mobo being though it is USB 1.1? I know this mobo has not been fully tested but since I have one here I would like to use it especially if there will not be any downside with transferring data, like movies, videos, and music.
Mark J. Foster 10-27-05, 11:15 AM Hi, Curtis104,
Since it's not fully tested, that means that you're willing to accept the possibility of lockups or data loss, and since it has USB 1.1, it will boot very slowly. If $75 is that important to you, give it a go.
Regards,
MarkF
ArchCarrier 10-27-05, 02:00 PM I have a RAID5 array, but now that I'm reading through this thread I fear that I've made the wrong decision about my storage. Is it possible to use my striped RAID5 hard drives with the UnRaid setup? I don't have a big budget, so I'd rather not buy too much new stuff.
Thomas J. Coyle 10-27-05, 03:24 PM ArchCarrier,
You can use your RAID5 hard drives with the UnRaid setup, but you will have a problem trying to use the striped drives directly. The UnRaid OS sets up each data drive with its own file system therefore there is no striping involved.
To convert to the UnRaid system you will have to go to normal ATA controller cards and somehow figure out a way to store the data that is on your striped drives until you have your UnRaid system setup and ready to store data. Hope this helps.
Regards,
TCIII
WIZD running. Joe's pointer to the mount point was all I needed. I've not yet kicked it off from a script but I believe that will be cake too.
To do this I DL'd FlipFlop's Windows port (http://www.geocities.com/flipflop7146/wizd.html) of WIZD, unzipped it, and copied the Skins and DocRoot directories to my array in a directory I'm reserving for Linux "stuff" along with the wizd.conf file. Next I copied over the WIZD Linux binary (http://www.geocities.com/flipflop7146/wizd.gz) from the same site. Telnet to the box, switch to /mnt/disk? (mine is disk2) Chmod this to be executable (775? cannot recall). Edit the WIZD.CONF file to match your setup. In my case I'm running it as console for troubleshooting and changed the name of the server to something I'd recognize. Lastly you'll need to setup a shortcut to get to your media - this had been tripping me up. However thanks to Joe it was easy - my media is on Disk1 in a directory named MP3 (case sensitive). So, my MP3 shortcut became /mnt/disk1/MP3 and I made sure there was no # on the start of that line. Fired up WIZD from the console via ./wizd and TaDa my LinkTheater sees it just fine :)
Not been playing long with it as I want to get YAC running and automate this launch but wanted to let folks know how I did with this. Checking memory with TOP I see 352megs used - I have a Gig in the box. I don't see WIZD listed so I cannot tell how much it's eating but if you've got just 512meg I'd be careful.
On to YAC :P
madpoet 10-27-05, 06:55 PM Excellent, thanks for blazing the trail
Hrm, up and working but for some reason it's not copying the file to the CRON job and WIZD isn't firing. I'm wondering if maybe the mount isn't available when that portion of the GO runs. I see no errors on the console <shrug> Firing it from Telnet is easy enough, I've not modded the check drive script to send me messages for UP drives but may do that to verify operation. Can provide more detail if anyone wants to try to figure this out etc.
WIZD working pretty well, has locked up my player a time or two but I suspect it's my MP3 source rather than WIZD or whatever lol.
Mark J. Foster 10-27-05, 09:43 PM Hi, Tom!
I'm ready to be your SATA guinea pig/first customer, if you're looking for one. Prices on the next-generation WD 400GB WD4000YR server drive (1.2M hour MTBF, 5 year warranty) have just crossed my pain threshhold - I'm ready to order a dozen of 'em. To do that, though, I'll need a box to put 'em in! If you're ready to go, I am! The part that might help convince you is that I also have a TeraStation that'll still be online during initial testing, so teething problems (including data loss) won't be a huge problem here initially. In addition, I'm perfectly comfortable if the drives don't spin down until a future release of the code.
What's the price for the 12-slot, no drive, dual power supply SATA version? If it helps, I've been running both Windows and Linux systems for a long time.
Thank you!
Mark
ArchCarrier,
You can use your RAID5 hard drives with the UnRaid setup, but you will have a problem trying to use the striped drives directly. The UnRaid OS sets up each data drive with its own file system therefore there is no striping involved.
To convert to the UnRaid system you will have to go to normal ATA controller cards and somehow figure out a way to store the data that is on your striped drives until you have your UnRaid system setup and ready to store data. Hope this helps.
Regards,
TCIIIAs Thomas has said your biggest challenge in re-using the hardware you currently have is that you must "break" your Raid5 array in order to use any of its disks in the unRaid server. To break it up you must have an alternative place to temporally store your data currently on it.
Now, once you "break" the Raid5, you can then re-format and reuse the disks on an unRaid server. This problems you are facing are some of the exact reasons I never configured a Raid array at home previously. The tasks involved in expanding it or adding disks usually meant needing a place to put the data while the upgrade was taking place and upgrading it meant all the disks needed to be upgraded at the same time to the same bigger size.
You might be able to "borrow" one drive from your current array and let it think it has failed and then use it in combination with another new disk to begin the unRaid array. You run the risk of a second disk failure in your current array and total loss of your data if you go about it that way, and you still need a place to put the balance of your data when you go to re-use the other disks in your current raid array so I'm not too sure I'd choose that route.
Thomas summed it up pretty well... you need to somehow figure out a way to store the data that is on your striped drives elsewhere until you have your UnRaid system setup and can re-use your existing drives in it.
How big is your current Raid array? How many writable DVD's would it take do do a full backup?
Joe L.
Mark J. Foster 10-27-05, 11:24 PM Hi, Tom!
No pressure or anything, but I just couldn't resist - my 4.8TB is enroute! ;) Pretty please?
Thank You!
MarkF
musicmann 10-28-05, 01:07 AM Joe L., I appreciate you trying to help ArchCarrier, but what I don't appreciate is you (inadvertently) reminding me why I'm still kicking myself. Honestly, Tom L-T, if your original post had come one month earlier, my life would have turned out totally differently! J/K of course, but Joe's post points out my big problem with homeusers and RAID. Either you pay upfront by getting an expensive RAID card that does OCE, or you pay upfront by having to max your system out from Day 1 to avoid the issue of not having OCE. Even though I got a great price on a 12-port 3Ware, I would have done things differently had unRaid been an option at the time.
MJF: That's just plain rude! Those of us with a lowly 2.5TB with about 600GB that you have to fill (but can't) before you can go unRaid are a little bit jealous.
Mark J. Foster 10-28-05, 07:45 AM Hi, Musicmann!
Honestly, Tom L-T, if your original post had come one month earlier, my life would have turned out totally differently!Ain't it the truth? Tom's created such a cool product that's it's a life-changing experience! :)MJF: That's just plain rude! Those of us with a lowly 2.5TB with about 600GB that you have to fill (but can't) before you can go unRaid are a little bit jealous.Sorry about that! ;) What finally knocked me over the edge was discovering the specs on these new drives (W.D.'s been working their tail off to improve quality after earlier problems), plus a shipped price of $220.32/400GB drive, from Monarch Computer (http://www.monarchcomputer.com/Merchant2/merchant.mv?Screen=PROD&Store_Code=M&Product_Code=150467&Category_Code=SATAHardDrives), which is 55 cents/GB. I can't wait! :p
Have Fun!
MarkF
Mark J. Foster 10-28-05, 08:07 AM Hi, Gang!
Just a few thoughts on why I chose the new Western Digital RE2 drives. The other drives I'd considered were Hitachi's high-end 500GB Desktar, and the Seagate 500GB Barracuda. The Hitachi drive draws almost as much power at idle (9.6W) as the W.D. does while actively reading/writing (10.0W, with an idle power of just 8.0W). While the Seagate was tempting, the 500 GB version has a very slow 11ms seek time. The Hitachi just edges out the W.D.'s 8.7ms seek time with its 8.5ms seek time, but the W.D. drive has a full 16 MB buffer, just like the 500GB drives. Obviously, the highest MTBF in the industry (rated at 1.2M hours with 100% duty cycle at a case temperature of 50C - these drives might even last longer if they're not spun down!), together with a 5 year warranty were the kicker. From my perspective, this is the best price/performance high capacity drive available right now, and since the OEM quantity 1 price is $250, I'm sure you'll see even better prices online soon!
Cheers!
MarkF
madpoet 10-28-05, 08:46 AM Thanks for the details Mark. I was wondering why you were so keen on that particular model.
Mark J. Foster 10-28-05, 09:19 AM Hi, Madpoet!
Thank you! I should have mentioned that "RE2" stands for RAID Edition 2. These drives are specifically designed for RAID-like applications, so don't run out and buy one for a desktop. Among the optimizations they've made are some tweaks to the motor or head control algorithms for vibration control in tightly-packed drive configurations, and they've also limited the amount of time that the drive will crank away trying over and over again to recover from a hard error (this prevents repeated controller errors from interfering with the RAID/Unraid data recovery process).
Since these drives also support Native Command Queueing (the SCSI-like ability to automatically reorder commands for optimal performance as the heads sweep back and forth across the disks when multiple applications are addressing the drive simultaneously), I wonder if the Unraid software will also support NCQ? It'll be fun to experiment with this and see!
Have Fun!
MarkF
Mark J. Foster 10-28-05, 09:44 AM Hi, Folks!
If you're curious about how MTBFs translate into the percentage likelihood that a specific drive or drives will fail over time, I found a delightful little white paper that describes it all in a very simple, easy to understand manner, particularly focused on hard drives. Check out Zzyzx Peripheral's MTBF and Availability Primer (http://www.zzyzx.com/products/whitepapers/pdf/MTBF_and_availability_primer.pdf). Recommended reading!
[EDIT] To tack on a simple example, in a 12-drive configuration of the RE2s, there's a 35.5% chance that at least one of the drives will fail over a five year period of time. This doesn't count items like controller failures, power supply failures, software bugs, etc, but it does provide a very useful gauge of how long things ought to last!
Have Fun!
MarkF
Hi, Folks!
If you're curious about how MTBFs translate into the percentage likelihood that a specific drive or drives will fail over time, I found a delightful little white paper that describes it all in a very simple, easy to understand manner, particularly focused on hard drives. Check out Zzyzx Peripheral's MTBF and Availability Primer (http://www.zzyzx.com/products/whitepapers/pdf/MTBF_and_availability_primer.pdf). Recommended reading!
[EDIT] To tack on a simple example, in a 12-drive configuration of the RE2s, there's a 35.5% chance that at least one of the drives will fail over a five year period of time. This doesn't count items like controller failures, power supply failures, software bugs, etc, but it does provide a very useful gauge of how long things ought to last!
Have Fun!
MarkFMark,
Great reading...
In the past 20 years or so I've personally have had about 8 drives fail in machines im my own home. Now, the reliability today is "probably" better, but I'll tell you the 30 meg drive in 1985 was a pretty expensive piece of hardware and having to replace one put a serious dent in the bank balance.
Now, since I usually have 3,4,5 or more PCs in the house I get to use the formula shown in the white paper to figure out the possibility of a single disk failure. I'm guessing in the past I averaged about one disk replaced every two years because it failed. If I have to replace one disk every 5 years in the unRaid array I figure I'll be way ahead. At least I won't lose the data on it. That was *always* a mess to recover from, even if I did backups on a regular basis it seems the failure occurs when least convenient.
Joe L.
Richard_P_Harvey 10-28-05, 12:19 PM From my perspective, this is the best price/performance high capacity drive available right now, and since the OEM quantity 1 price is $250, I'm sure you'll see even better prices online soon!
Cheers!
MarkF
Mark,
Where are you able to find the 500GB RE2 at $250....?
Tom, the UPS in question is an APC Smart UPS 900 (might be 9K). It has a serial port on the back, I'm taking it home tonight...
Thomas J. Coyle 10-28-05, 01:29 PM BLKMGK,
I presently have my DVD server on a 1050 watt APC UPS (Costco had them on sale for $99.95 a while back) with line voltage regulation and has a USB port on it for communication with the attached PC.
Regards,
TCIII
Mark J. Foster 10-28-05, 01:36 PM Hi, Gang!
J.L.: I'm glad you liked it! You and I are obviously in similar positions. I've got something like nine PCs running around here at the moment, and have had plenty of experience with drives dying. In particular, despite having built up eight different Shuttle machines, those boxes always run the drives too hot. That's one thing that I didn't mention in my post above, but running the drives hotter will dramatically decrease the MTBF. I believe at this stage that I've lost four different drives, all from different manufacturers, out of the six Shuttles that I've run here. Yuck! Having a really big storage server around to ensure that everything is always frequently backed up is going to make my life a lot more pleasant! Heh, thank you for the nifty YAK script!
Richard: Sorry, you missed the earlier post. These are 400GB SATA drives - W.D. doesn't make a 500GB drive yet!
Cheers!
MarkF
Wow, nice sale! I missed out it seems. This one has been upgraded with some better batteries by a friend of mine - should do me well. Tom had mentioned supporting UPS at one time so I'm hoping this one will be easy to support - we'll see. If not no sweat, it's better than nothing and it will catch the normal transients we get here. Thankfully I seldom lose power for very long.
Mark, I run Shuttle XPCs too but so far no drive failures. Mine don't seem to get tooo hot. Actually though come to think of it one of them has weird issues. Torrents would always fail CRC check and the OS slowly died on me. I wonder if it was heat but if so that was the only box doing that. That drive may end up in my array soon enough so we'll see if that drive was bad or not. I could never get that system stable! All of my other XPCs are running well inclduing those built for friends. However I also echo your feelings on backups - NONE of my systems are backup right now because I simply have never had enough space to do so. I will now :) I'll be using Acronis and scheduling backups I think - will be a big relief to get this done...
pischke 10-28-05, 01:58 PM BTW the Cremax mobile racks can be had for under $20/each.
OK, I'll bite - where might one find these for < $20? Any sources you know of in the SF Bay area?
Thanks.
-pischke
pischke 10-28-05, 05:56 PM One other question Tom - is there any way to have drives mounted as read only? Or at least shared out as read only? Once a drive is full, it would be handy to be able to set it as read-only to avoid "accidents".
-pischke
(BTW, I'm loving this product! For once, my tendency to procrastinate has served me well, as I had been meaning to put together a raid 5 box for quite a while, until I saw this thread)
Mark J. Foster 10-28-05, 09:23 PM Hi, BLKMGK!
While I apologize for the off-topic post, over the years I've learned that there are three contributing factors to the high drive failure rate in Shuttles. First is running them 24 hours/day, which I do with all my PCs. What happens when you do this is that the otherwise-nifty I.C.E. cooling module's heatsink gets gummed up quite quickly, and temps soar.
The second problem is if you try to quiet your system down via the Smart BIOS setting, you'll regret it! The problem is that the BIOS pays too much attention to the CPU, and not enough to the drives, and the drives will tend to overheat.
The third factor is actually a variation of the second factor. When you run applications that really bang on the disk, the probability of a drive failure grows very rapidly. I've had two drives fail while running the PerfectDisk disk defragmenter. A third drive failed while running the Windows XP Service Pack 2 upgrade, for the same reason.
That "weird" Shuttle you have sounds like it, too, may well have been a victim of the drive overheat problem in Shuttles.
Meanwhile, back to your regularly scheduled program! Sorry about the diversion!
Cheers!
MarkF
Mark J. Foster 10-29-05, 10:39 AM Hi, Tom!
It just occurred to me that my mentioning NCQ might be translated as a request for that functionality over schedule. If so, please ignore it! If your solution supports NCQ, that's great, but I'm much more interested in moving forward with an SATA box than I am trying to make anything more complicated, or delaying the release. So, please ignore this issue if it'll get in the way!
Cheers!
MarkF
Thomas J. Coyle 10-29-05, 05:07 PM J. L.,
The UnRaid OS webpage is showing 12 drives (1 parity + 11 data), however My Network Places is only showing 10 data drives. Do you know where to look in the Flash Drive files to determine what the OS is sending out to my Windows XP machines? I sent an email to Tom, but there has been no response yet.
Regards,
TCIII
limetech 10-30-05, 12:23 AM ... Lime Tech might be crazy enough to violate the GPL, but I don't think they'd be crazy enough to attempt a backport of reiser4 to the 2.4 kernel :eek:
* One may only be excused from violating the GPL on the grounds of insanity think RMS said this once];
* One must assert one's insanity to be excused on this basis;
* One who requests to be excused is presumably in fear for his business. This is taken to be proof of his sanity, and he is therefore obliged to continue adhering to the GPL;
* One who is truly insane presumably would not make the request. He therefore would continue following the GPL, even though as an insane person he could of course be excused from it simply by asking.
Heck, I still wouldn't be totally comfortable with reiser4 on a 2.6 kernel. It's still very much in development...
reiser4 or the 2.6 kernel?? [I think both - which is why we're using neither]
Actually, now that I think about it, does the parity drive have a file system on it as well? I would guess not, since from info gathered here (if only www.lime-technology.com had half the info this thread does ;) ) the parity drive can't be any arbitrary drive, ie. it has to be the first drive in the array (/dev/hda). And since the XOR is done in kernel space, it would be much more efficient to do I/O at the block level...
There's no file system on the parity drive - it's strictly block-level parity of the drives in the array. Also it does not have to be any particular drive. We simply chose it to be the "first drive" for simplicity. You can modify the "go" script on the Flash to designate any of the slots as the parity drive.
limetech 10-30-05, 12:36 AM WIZD running. Joe's pointer to the mount point was all I needed. I've not yet kicked it off from a script but I believe that will be cake too.
To do this I DL'd FlipFlop's Windows port (http://www.geocities.com/flipflop7146/wizd.html) of WIZD, unzipped it, and copied the Skins and DocRoot directories to my array in a directory I'm reserving for Linux "stuff" along with the wizd.conf file. Next I copied over the WIZD Linux binary (http://www.geocities.com/flipflop7146/wizd.gz) from the same site. Telnet to the box, switch to /mnt/disk? (mine is disk2) Chmod this to be executable (775? cannot recall). Edit the WIZD.CONF file to match your setup. In my case I'm running it as console for troubleshooting and changed the name of the server to something I'd recognize. Lastly you'll need to setup a shortcut to get to your media - this had been tripping me up. However thanks to Joe it was easy - my media is on Disk1 in a directory named MP3 (case sensitive). So, my MP3 shortcut became /mnt/disk1/MP3 and I made sure there was no # on the start of that line. Fired up WIZD from the console via ./wizd and TaDa my LinkTheater sees it just fine :)
Cool! :cool: I don't know much about the LinkTheater - can you set it up to be able to see media on multiple shares? How about if all disks were under a single share? I think we'll go buy a LinkTheater just to test this out.
Not been playing long with it as I want to get YAC running and automate this launch but wanted to let folks know how I did with this. Checking memory with TOP I see 352megs used - I have a Gig in the box. I don't see WIZD listed so I cannot tell how much it's eating but if you've got just 512meg I'd be careful.
Caution: don't rely on 'top' to give you accurate memory usage information. There is a lot of memory shown as 'free' by 'top' which actually isn't. Also do not try to run the UnRAID s/w on less than 512MB.
limetech 10-30-05, 12:50 AM Hi, Tom!
I'm ready to be your SATA guinea pig/first customer, if you're looking for one. Prices on the next-generation WD 400GB WD4000YR server drive (1.2M hour MTBF, 5 year warranty) have just crossed my pain threshhold - I'm ready to order a dozen of 'em. To do that, though, I'll need a box to put 'em in! If you're ready to go, I am! The part that might help convince you is that I also have a TeraStation that'll still be online during initial testing, so teething problems (including data loss) won't be a huge problem here initially. In addition, I'm perfectly comfortable if the drives don't spin down until a future release of the code.
What's the price for the 12-slot, no drive, dual power supply SATA version? If it helps, I've been running both Windows and Linux systems for a long time.
Thank you!
Mark
For SATA, we're going to be releasing a 12-drive tower version (same CM Stacker case), and a 12-drive 2U rackmount version!
|
|