Four drives: RZ2 or RZ1+backup?

[mappo]

I'm buying four 4TB WD Reds for my personal data.

I can't decide whether to put
A. all four in an RZ2 vdev - or
B. three of them in an RZ1 vdev leaving the fourth as local backup.

Would option B entail a 2nd 1-drive pool?

Is there a performance penalty for multiple pools/vdevs?

I'm already planning on a separate pool for a BT jail (probably Transmission though I'm open for suggestions) utilizing two desktop-grade 2TB drives that already have a couple of years in them. If they fail, I can live with losing BT data but I don't want them to bring down my personal data.

 

[darkwarrior]

Hi there,
personally, I would go with RAIDZ2 on 4 drives, because of the size of them and the potential URE issues encountered during a resilver.
Option A is offering you redundancy against to 2 drive failure, but you will still need a backup...

There would be an alternative:
Creating a pool of striped mirrors with 2 * 4TB mirrors + 1*2TB mirror.
That would offer roughly 9.2TB of diskspace (- 20°% percent of course)

But if you go with your option B you are not forced to have 2nd local pool.
You could connect it via USB to the local machine (or another PC on the network ) and copy/paste the data to it.

Have fun ;)

 

[depasseg]

When you say "backup", do you really mean backup, or do you mean hot spare?

I wouldn't worry about a hot spare, but the backup drive is an interesting dilemma. I think it would depend on drive sizes. If the 3xRaidZ drives were 2TB or smaller, I'd go with a 3 drive RaidZ1 + Backup. If larger than 2TB, I'd go 4 drive RaidZ2.

 

[Scharbag]

If you are buying 4 4TB drives, then you have only 2 options IMHO:

1. pool made of one 4 drive RaidZ2 vDev (my preferred method)
   • this offers excellent redundancy and the performance will be more than good enough for home use
   • can tolerate ANY 2 drive failures without data loss
2. pool made of two 2 drive mirror vDevs
   • this will offer slightly better performance than a single vDev
   • theoretically can still handle 2 drive failures as long as they are in separate vDevs (one reason I prefer RaidZ2)

I would avoid the RaidZ1+spare option due to the risk of failure during resilvering as RaidZ2 has the exact same amount of overhead (50%) as this option and offers better fault tolerance.

My assumption was wrong :) I would still go with Z2 and then follow up with a good backup strategy.

Cheers,

 

[mappo]

@depasseg : Thank you for answering! I'm talking about an actual backup - not a spare. And it's not a mirrored backup - just a simple cron job "copy these files and directories to this separate disk once every night"

Of course I could do what @darkwarrior suggests and keep it on a separate host or on usb, and it might come to that if there are issues with an "in-case" backup disk.

There will be an off-site backup as well; a weekly replication of the internal backup disk, probably a RPi at a friend's house or something.

 

[Robert Trevellyan]

Would option B entail a 2nd 1-drive pool?

Yes, according to your statement that you did not mean "hot spare".

Is there a performance penalty for multiple pools/vdevs?

Possibly. When I had 2 pools, my ARC hit rate dropped.

 

[mappo]

Possibly. When I had 2 pools, my ARC hit rate dropped.

Thanks ... I think. I didn't understand the latter part of that sentence. What wizardry is this this ARC hit rate of which you talk?
More cpu load? Longer access times? Laggy network? Dandruff?

Is it reasonable to think I'll avoid the penalty by hooking the backup drive up to usb instead?
Will the price for three pools mean even higher that for two?

 

[Robert Trevellyan]

What wizardry is this this ARC hit rate of which you talk?

Start with this: http://doc.freenas.org/9.10/zfsprimer.html#zfs-primer

Is it reasonable to think I'll avoid the penalty by hooking the backup drive up to usb instead?

Not unless that drive is detached most of the time.

Will the price for three pools mean even higher that for two?

I don't have a good answer for you, I just reported what I observed on my system, unencumbered by deep knowledge of the mechanism that led to it.

 

[Robert Trevellyan]

A. all four in an RZ2 vdev - or
B. three of them in an RZ1 vdev leaving the fourth as local backup.

Coming back to your original question and the title of this thread, I'll reiterate what everyone must recognize about RAID, which is that it is not a substitute for backup. No matter what pool/vdev configuration you choose, you must backup any data you care about.

EDIT: I see you are planning offsite backup. Good.