A first look at NZFS and replacing unRAID with NZFS’s Transparent RAID (tRAID) - Page 4

I built my server a few years ago. I remember having some difficulty at the time. I think there weren't that many people building new servers at the time. Most of the options that met what I was looking for had an on-board network adapter that was problematic. I ended up just buying something and hoping it would work.

It's not necessarily that easy just to copy builds. For a lot of things, there's someone that thinks there's a problem. The SIL3132 cards are a good example of that. They're basically the closest thing to the recommended 2-port SATA card. But some people seem to be having problems with it.

Maybe a Windows-based server would run into the same problems. But, as I said before, I suspect the drivers/firmwares are better tested on Windows than Linux. I'd probably give it a shot if there was a good unRAID alternative on Windows.

I currently have an older thecus I'm looking to replace before it becomes an issue in the very near future. All I want to know is how I can become a beta tester for this platform!

if you take a closer look on the wiki entry for Sil3132, you will notice that the second link points to my problem solving thread (June 2012). And the first one points to a thread (February 2012) which was linked in my thread two days into problem solving (see here). The Wiki entry was updated (to my best knowledge) with a warning only after I had reported identical problems. When I bought the second 2-port Sil3132 based card I had already forgotten anything about the controller it was using, I was simply buying something identical what I had previously analysed and proven in production to be 100% ok. Normally I can find really quickly similar problems and any existing solutions but in this case I was simply missing the magic 3132 search term. From my problem solving thread linked above you can see that I did spend quite some time trying to narrow down the problem, including producing re-usable scripts. I think it was closer to 15 hours of work time in total which I spent switching cards/drives, updating bios and bios settings and performing scripted tests. As a contrast I remember it taking about 1-2 hours to install the Supermicro AOC-SASLP-MV8 and upgrade from 4.4.2 to 4.7.
I was running 4.4.2 from 2009 until last year with no problems. I was forced to update to get support for AOC-SASLP-MV8 which is in 4.7 (for 2TB discs). I'm not really fond of an idea to use beta version in a production storage system. But now I have a need for both 3TB discs and running Plex Media Server on UnRaid which are only supported in later 5.0 release candidates.

To be clear, I think UnRaid was/is the most perfect solution to my needs. I took a short look back in 2009 on the alternatives and considered UnRaid's feature set to be superior to any other solution. I have followed only UnRaid forums and even those only when I have had a need to change or add something. But now I'm closing max capacity on my first system and I'm looking for the next thing. From HW point of view it's easy, it will be a rack mounted 24 drives case with SuperMicro stuff inside Just lacking the software.

Are we trouble shooting unraid in a flexraid or NZFS thread? Seems Ironic.

Here I thought it was a nzfs vs unraid thread... Which there isn't much to say about nzfs other than that it has the exact same advantages that flexraid has, same disadvantages too!

What is the disadvantage of flex or Nzfs ?

I didn't check the links so I never saw your thread. If you were involved in uncovering the issue then kudos to you for getting to the bottom of it.
I'm fairly sure that unRAID had 2TB support prior to version 4.7. I've been using a 2TB parity drive for at least a couple of years and I was definitely using an older version of unRAID. I'm also using two of the AOC-SASALP-MV8 controllers in my server. However, I've got the parity drive connected directly to the motherboard and not the controller. I recently added a 2nd 2TB drive as a data drive that is connected to one of the SATA controllers, but that was after I switched over to 5.0-rc8.

FWIW, version 5.0-rc is a release candidate and not a beta. Apparently it's gone through a multitude of changes as a release candidate so it's anybody's guess when it will become an official release. The main reason I switched over was for the additional drive support. It's worked pretty well so far and I've been quite pleased with it. The UI is so much better than the pre-rc 4.X versions.

To be fair, even plug and play solutions have limitations (e.g. not all NAS have 3TB support), and in those cases, the solution is to simply buy a new device.

If someone is not technically proficient, it is less likely they will have a need for more than 6 drives, in which case dealing with addon sata cards is not common and not really a good example for advising against software RAID for non technical people past the initial setup. In any case, those non technical people can also buy ready made solutions (at least in the case of unRAID). It's only the more technically adventurous that tend to find ourselves in a spot of bother when we try to add extra doodads and what not to a basic device.

For a very long time I used unRAID as a pure file server (no add-ons - just the basic 4.7 release), It performed flawlessly for that role. That is the key point to take away - different tools for different jobs. unRAID was never really marketed as being an all in one solution, it was only due to the enthusiastic support from tech savvy users that it's functionality began to expand. This functionality tempted a few others to give it a try and so we have a situation where people are so reliant/used to their add-ons that they forget to disable them when beta testing the release candidates all too often.

Wow. I will never support you or your products. I hope your superior product is heavily pirated. Well not heavily, because I doubt anyone will know what the product is with such a stupid name.
Edited by hemogoblin - 2/17/13 at 7:41pm

It's so refreshing to read someone post this. People sling all these blanket statements about what's the best before ever asking what someone wants to actually accomplish. Bravo.

...
Edited by PobjoySpecial - 5/16/13 at 2:30pm

Precisely. The facts as they apply to this forum are that most people here use servers to distribute media throughout their homes and not primarily as data repositories and backups. This is why I feel that the OP is aiming his product at the wrong market by announcing it here.

Bring your own operating system.

Takes longer to build.

Runs off of hdd/sdd. Unraid runs in memory.

Windows.

Take down array to initialize drives(not sure this one is true). Unraid I can preclear drives and then add them, Stop, Add, Start(<1 minute down)

Developer seems like a turd. Support in future versions of windows for flexraid? Sounds like he is moving on. Unraid I can update myself for future versions of linux.

Unraid/Flexraid/NZFS are all in essence the same thing.

I'm not sure I agree. I like added features of WHS so I see it as a superior solution.

Windows doesn't bother me.
Developer doesn't bother me.

Windows has advantages along with its disadvantages. I'm really just saying that there are disadvantages.

The developer now has 2 pieces of software to develop/support. The new one will most likely be more expensive. Which will get more attention?

If UnraidZ came out and development of Unraid slowed or stopped, I would be mad.

I have run flexraid for over a year at this point and am super excited about this.

Windows doesnt bother me either. In fact being able to run more services on a single box is great. Saves time/electricity/space.
My main HTPC/NAS/Plex/XBMC/MySQL/VMware/Itunes/Couchpotato/SABNZB/SickBeard is all on one box that runs every day from 8am to 1am.
This saves me a ton on electricity over having a seperate NAS and HTPC. The only other option that would be cheaper would be a NAS that used WOL when it was needed.
But this leads to problems when using always on services anyways. So windows/flexraid is an awesome combo for this.

I also have no problem with the Developer and was able to get help when I needed it.

How often do you need to add drives. If you are afraid of the ten minutes of downtime needed to add a drive you need to look at enterprise class SANs.

Not to mention flexraids ability to use any type of drive local or otherwise. Oh and using NTFS and not worrying about an odd format is nice.

My htpc sleeps, my server(NAS) is on, idling. I'm fairly certain the difference in power usage between our setups is at most a couple dollars a year.

Most of the services you are running can be added to unraid really easily. I wouldn't really suggest you buy unraid if you already bought flexraid(they are the same thing).

I would never suggest using a network drive in a parity based array, if your network goes down your parity is invalid?

I believe the long initialization with flexraid is when you add a full drive.

Reiserfs really isn't an "odd" format.

To each their own.
I used unraid for about two months but ended up blowing it away when I realized could run all the services on a single box. No need to worry about sleep/wol was nice as I have two secondary htpc's (who do use WOL) that also pull their tuners/content from the main box.
I also had sleep issues with the IBM m1015 controllers with unraid it was probably a driver issue of some sort.
I gave my unraid license to a friend who still uses it.
I agree with the network drive parity issue just saying its possible

For me Flexraid was the best fit with the hardware/needs I had. I am eager to see what comes of this in the future.

Well..

Here is what I have learned on the matter.

If your making a simple flexraid array with empty drives it takes almost no time. Example:
4 HDD's, 3 storage and 1 Parity all empty. 3TB each. Would give 9TB usable space.
If all drives empty it takes no time at all.

If your making the same 4 HDD array but 3 of the storage HDD's are full it takes about 10 hours to do it. Put it on a night and it's should be done in the morning.

Flexraid speed it tells you is based on the speed of the slowest drive in your array. Usually for me it's a 2TB green drive and about 85MB/sec

But,

As explained in the thread I had on the subject the reality is it's processing all your HDD and reading at the same time. So it's really reading at 85MB/sec x 3 or there abouts. Meaning it is computing data at about 250MB/sec+ which is why it takes much less time than I was figuring it to take initially.

I did a 20TB array in the same time I did a 9TB array. That made no sense to me. But here is how it works.

I did 6 HDD's of 3TB each for 18TB and then one 2TB green drive. Used a 3TB green drive as my parity (only mention green drives because they are slow giving a worst possible case scenario)

It was reading basically at 7 x 85MB/sec = 595MB/sec

So in reality it takes only as long as it takes to read all the data from your most full or slowest drive- regardless if you have more or less data.

That is why 20TB @ about 595MB/sec takes about the same time as 9TB and only 250MB/sec. In reality it's the same process simultaneously from all drives. reading and computing parity data.
On any modern CPU the CPU or computation process is not the bottle neck. It is the speed of the HDD's. So the time it will take depends on the speed your HDD can be read, and also how much data needs to be read (how full your drives are)
The amount of time it will take to read all data from your HDD is how long it takes. Since it reads them all at the same time, whichever drive is the slowest/most full is the time it takes.

10 hours to initialize 20TB worth of full drives is really actually pretty good performance when you consider what your doing.

I understand the benefit of the new NZFS is going to be much less time though.

...
Edited by PobjoySpecial - 5/16/13 at 2:30pm

flexraid only reads the drives. It does not erase them. The only drive that has data at all changed is the parity. That drive and it's content are completely erased so you know this before you set it up and chose a drive knowing that.

Even if flexraid failed initialization or your new parity crapped out in the process of setting up you would not lose any data you had on data drives. Once it's up and running your data is protected with parity back up.

I am not sure your fears are realistic. You can't expect flexraid to offer data protection before you have it set up. That makes no sense.

When you add a new drive to unraid there is no parity calculation.

If you are writing to the array and a drive fails, it will continue the parity calculation without that drive by using a "virtual drive"(kind of). Probably not a good explanation. It would calculate what is on the failed drive through parity and then rewrite the parity with this new information.

^^^If that is nonsense please let me know.

Basically you are always protected with Unraid once you initially build the array.

Please correct me if I'm wrong but when you add a full drive to Flexraid you are rebuilding the array from scratch aka initialization and your data is not protected. This could be consider a bad idea and a bad feature. This is nearly as bad of an idea as adding a network drive.

...
Edited by PobjoySpecial - 5/16/13 at 2:29pm

1) I had a drive fail during a write and I didn't know it failed until I got an email saying so. Lost no data.

2) I believe I know the answer but someone with Flexraid should answer this. I don't think you will like the answer.

I think your concern is overblown here.


First, It would not matter if the drive is full or empty in either case. It's the same. Flexraid just reads the data. If it's empty it just has less reading to do than if it is full. But the result and reality is the same.

Second,

If your adding a drive in flexraid (full or empty) then your going through the initialization process and creating your parity. So yes during this short time technically flexraid is not running yet so if a drive failed you would not be protected. But this is a short window of time (only hours) and as soon as initialization was done you'd be ok again. I think it's just very unlikely you'd have a data drive die in that short window of time. If you had serious concern about this happening the solution would be to use the new HDD for parity on the new initialization- or simply just use a new drive as your parity drive and preserve the old one. You could remove the old and original parity drive and/or just take it offline during the new initialization. That way if flexraid did not complete initializing and drive actually failed during the process you could bring back online the original parity drive and rebuild it. Seems like a huge waste of time and paranoid actions to me- but that is what a paranoid person would need to do.

Bottom line is - if flexraid is not running or set up then your not protected by it's benefits. All it's going to do is read your data from your drives. But if you really wanted to make sure you did not have a data drive die during that time you'd need dual parity drives.

Your concern basically goes along the lines of the same questions regarding simultaneous failures. Some people always ask what if I had two drives die at the exact same time. Then the answer is you would need two parity drives to allow for dual and simultaneous failure of 2 HDD's. For smaller arrays in home use- Most only use 1 Parity drive. Beyond 8 HDD's I think you should probably be using 2 parity, but I have seen some people run dual parity drives with only 2 or 3 data drives. Seems crazy to me but hey... who am I to judge.

The advantage of Flexraid is you get to back up with parity multiple HDD's with only a single parity backup drive. This saves you on the cost of additional HDD's. Backing up 4 HDD's with only 1 HDD saves you the cost of the missing 3 HDD's versus needing 4 HDD's to back up 4 HDD's with duplicate copies. 3 HDD = $300 or so saved on drives. Makes good sense and value for most.

Your line of questioning makes me wonder how at all you'd even want software raid in that case. If you were really concerned with such an unlikely event occurring you'd be better off going 1 to 1 duplication so you could tolerate all your drives dying at the same time.

I'm trying to help answer but your concerns are over exaggerated IMO.

I think the benefit with Flexraid of having your data and files on your data drives be readable and copy-able easily and instantly from any machine including USB or a HDD dock station far out weights the concern over a drive failing during initialization and not being protected. It's worth the trade off to me.

Your question is kinda like this question:

"I am thinking about Flexraid but have not bought it yet, but I plan to buy and set it up very soon; What if one of my drives dies before I buy flexraid and install it- Is my drive and data still protected by flexraid ? "

It's just not making much sense to me.

If you use Flexraid in the same way you would use Unraid then your data is protected. If you use it in a different way then it might not be.

So Flexraid supports pre-clearing and burn-in testing? And when used, the addition of the new drive to the array takes place in few minutes without the need to write to the parity drive? Since this is how it works on Unraid (though needs the 3rd party pre-clearing script from Joe L.).

Unless I'm misunderstanding the scenario, it's nothing like having two failures at once.

As I understand it, when you add a drive in FlexRAID, you lose parity protection until you can recalculate parity. At this point, one drive failure is all it takes to lose data. Now I agree that its not the worst thing in the word to lose parity protection for a while, but in many ways this is a terrible time to lose parity protection. You're about to go through all the data on all your drives at once. If there's going to be a mechanical failure on a drive, that's a pretty good situation for it to happen. These are typically setups where the drives are usually sitting idle. It's during these parity calculations that they really get a workout.

That's not the trade-off. The trade-off is that you get to add drives with data on them to the array. unRAID has the same property of being able to read individual drives. unRAID doesn't lose parity protection when you add a drive, but that's because drives are wiped when they're added to the array. You do, however, lose parity when you decommission a drive. But that's just because Tom has never bothered to write a "safely remove drive" method.