ZFS Best setup for backup server disks

R

Reaperzx

I have following HDD-s in backup server: 2x 16TB and 6x 6TB. What would be the best way to make one big disk of them? Using FreeBSD 15.1.

(Background info: I upgraded my backup server CPU and at the same time cleaned it. Also yanked all HDD-s out and tested them one-by-one in Windows machine, because I have proper tools for that there. All disks work fine after that, except 16TB gmirror was gone. I have accepted by now that data is gone. It is not big problem, because it is offsite backup server. I though I might use this chance to change the disk plan).

My choices are:

1. Use it as was before upgrade/cleanup:
All disks were paired in gmirrors. And then big gconcat of gmirrors. Basically RAID 10 or more like RAID 1 + JBOD.
Worked fine so far. Around 32TB usable space.

2. Try ZFS. I still dont understand ZFS, but I guess I could make something up in AI.
I was surprised to learn that ZFS does not like different size of disks either. So how does this plan sound:
Pair 6TB disks together to make 12TB virtual disks (using either gstripe or gconcat).
Then I would have 3x 12TB virtual disks and 2x 16TB disks. I would use those 5 disks in ZFS RAIDZ2 and have around 32TB usable space.
Only 8TB space at end of big disks would be wasted.
 
From my understanding, if your ultimate goal is to use the ZFS for your filesystem, then don't gmirror nor gconcat anything. Straight ZFS only, I believe that ZFS works best when it has as much direct control of your disks as possible. It is a filesystem AND volume manager rolled into one, which is why it's so good.
 
VladiBG said:
2x16TB in zfs RAID1 = 16TB ~14TiB usable space
6x6TB in RAIDZ2 = 24TB ~21TiB usable space
Click to expand...
And then should i merge them together with gconcat?

Jakflip said:
From my understanding, if your ultimate goal is to use the ZFS for your filesystem, then don't gmirror nor gconcat anything. Straight ZFS only, I believe that ZFS works best when it has as much direct control of your disks as possible. It is a filesystem AND volume manager rolled into one, which is why it's so good.
Click to expand...
It does not have to be ZFS, frankly I don't care what file system it is. Just was considering ZFS because I have never used it before. Have been using UFS fine so far.

I just want to have one big disk for offsite backup. My experience with gmirror/gstripe/gconcat has generally been good so far. Less complicated to use, than graid3 and graid5.

If any of the disks breaks, I will try not to buy replacements in this economy. Probably will grab some old 3TB disks from the shelf and gconcat them together... 2x3TB = 6TB

Can I use gconcat as a member of RAIDZ2? AI said yes, but not recommended...
Relevant discussion in reddit:

I used Areca hardware RAID6 for years, but it did not support different size disks. Upgrading whole array at once was quite costly.
 
Reaperzx said:
And then should i merge them together with gconcat?
Click to expand...
no. You should not mix different size of disks. In any option you will have 2 raid volumes one from the 2x16tb in raid1 and one from 6x6tb in the raid setup which you choose.

The best option is to use hardware raid with bbwc in raid6 (~24TB) or if you don't need so much space and need more reliable option then use raid10 (~18TB) with UFS file system.
 
Reaperzx said:
I have following HDD-s in backup server: 2x 16TB and 6x 6TB. What would be the best way to make one big disk of them? Using FreeBSD 15.1.
Click to expand...

I would make 6 TB partitions on the 16 TB disks and do a 8-disk raidz2 or raidz3, and then a mirror on the 2x10 TB partitions.
 
You must log in or register to reply here.
Back
Top