Pratical Upper Limits Harddrive Size Limits

[Andrew076]

I seem to recall seeing this somewhere but can't find it now. It was a discussion about the practical upper size limits of the individual drives in a zpool. The discussion went something along the lines of the following:

If the individual drives were too large and you had a failure then the time it would take to rebuild the VDev could be too long and intensive you could suffer a second or third failure and then not be able to rebuild the VDev and could loose the zpool.

The reason I am asking is that I purposely bought 6 WD Red 3 TB drives with the plan to every three months buy a new disk. The "Plan" was to continue to do this indefinitely so as to help prevent several multiple drive failures where I couldn't recover the VDev. I have been running for 3 months now and am ready to buy the first drive and wanted it to be at the upper practical limit but I can't seem to find that discussion so to the extent someone could help me out I would appreciate it.

 

[cyberjock]

I'm confused.. what do you plan to do with this one disk? Replace an existing disk?

 

[Andrew076]

I'm confused.. what do you plan to do with this one disk? Replace an existing disk?

Yes, replace an existing disk... my assumptions were:

1) This would be the auto expanding option (when all are replaced more capacity)
2) Staggered like this would put them defiantly in different product runs and would reduce the chances of a multiple simultaneous hard drive failure (even though I can withstand 2 right now)

 

[Bidule0hm]

It's true for RAID-Z1 but as you use RAID-Z2 it's not a problem for you. If you want more details on the subject you can read the link in my signature about RAID-Z1/RAID5 ;)

 

[Andrew076]

It's true for RAID-Z1 but as you use RAID-Z2 it's not a problem for you. If you want more details on the subject you can read the link in my signature about RAID-Z1/RAID5 ;)

Thanks I am trying to be as conservative as reasonably possible. Before Freenas I had a Buffalo Terrastation that had two simultaneous (or near simultaneous failures) and it cost me $4K to get fixed...

So with RAID-Z2 and the fact that it can support two simultaneous failures is the method I mentioned above not really needed? At what size would it be?

 

[Bidule0hm]

Basically unlimited because it's the fact that you don't have any redundance with RAID-Z1 when you have one disk offline that is risky.

What you want is to do a good burn-in of the system (there's a sticky about that in the hardware section) to eliminate any infant mortality ;)

 

[toadman]

I haven't redone the calculations in a while, but I think you would be ok with RAIDZ2 at 3TB. If you you want to be conservative about it go RAIDZ3 and rotate. That would keep you good well beyond the practical life of the server.

I'm curious, what is your backup methodology?

I use mirrors on my mail pool because I'm hosting VMs. I backup to a different server with a RAIDZ2 pool of 3TB drives. The latter spin down except for the once a day replication.

 

[Andrew076]

I haven't redone the calculations in a while, but I think you would be ok with RAIDZ2 at 3TB. If you you want to be conservative about it go RAIDZ3 and rotate. That would keep you good well beyond the practical life of the server.

I'm curious, what is your backup methodology?

I use mirrors on my mail pool because I'm hosting VMs. I backup to a different server with a RAIDZ2 pool of 3TB drives. The latter spin down except for the once a day replication.

I am not sure really of what you mean. I think you are saying that you are mirroring each of the hard drives in your main pool and then back up to a separate server? That is hard core and way beyond me I think....

I am using the FreeNAS server for pictures and home videos (my wife takes a lot of pictures of the little ones and never saw a blurry picture she didn't think the technology would someday advance where it could be cleaned up and would have just that one shot that she needed... never mind the fact that she will never be able to find that one picture among the thousands she takes... but I digress).

To answer your question, for backup I am currently using CrashPlan as the backup. It is set to check once a day for any changes but what I generally do is when I upload a batch of pictures / home videos (which are kept redundantly prior to this) I have CrashPlan start the backup then. That way in about 6 hours I have a complete backup on CrashPlan again with the new pictures.

Having said that I have not had to use CrashPlan yet for recovery (and would like to never have to do that...!) hence the nervousness and desire to protect the Zpool. A second redundant backup is done by my wife (who loves her pictures) she also backs them up to snapfish as well.... so we could always prints of the pictures in the future. That doesn't do anything however for the home videos.

 

[cyberjock]

I'm not a fan of replacing drives "just to replace them". If they are excessively old (maybe 5+ years old), that's one thing. But new drives have infant mortality of their own.. their failure rates the first year are excessively high compared to years 2-5. So I wouldn't go replacing disks just because you can. If they fail, if they are >5 years old, if you need to expand, sure. But not to avoid failures. You'll probably see more failures that way.

 

[cyberjock]

Also there is no amount of ZFS redundancy that counteract's the need for a backup. You should always have backups, no matter how conservative you want to go with the live data.

 

[toadman]

Yes. My vdevs are 2 disk mirrors. 4 vdevs, about to be 5. The backup pool is a single 5 disk RAIDZ2 pool. (which I'm about to blow away and remake into a 6 disk RAIDZ2 pool.)

I'm not sure it's hard core really. Just a pool and a backup. In your case you have a pool and a backup (Crashplan) too.

As long as you have a reliable backup, I wouldn't worry too much about your RAIDZ2 pool. Really. Yes it would be a pain to restore from backup, but it's not likely to happen. I don't think you need to spend the money on the rotation. If anything, just do a RAIDZ3 (an extra $150?) and then not rotate (which is going to cost you $150 every three months right?). You'll be fine.

 

[toadman]

I'm not a fan of replacing drives "just to replace them". If they are excessively old (maybe 5+ years old), that's one thing. But new drives have infant mortality of their own.. their failure rates the first year are excessively high compared to years 2-5. So I wouldn't go replacing disks just because you can. If they fail, if they are >5 years old, if you need to expand, sure. But not to avoid failures. You'll probably see more failures that way.

Yep, I agree. For the stated use case it's not necessary.