Storage options

[EFW]

Looking to build a media server with TrueNAS. I would have 1 pool with 3 or 4 HDDs to start and then expand the pool when finances allow. Have I got the calculations, correct? I would also have 2x 500gb SSD for OS / Cache. I’m aware of storage overheads, these are just rough.

Each HDD is 18tb

ZFS1 pool - 3 HDD x 18 = 54 (54 - 18 = 36tb)
OR
ZFS1 pool - 4 HDD x 18 = 72 (72 - 18 = 54tb)

(3x 18 = 54)
(54 - 18 = 36)
(Par drive = 18 (18 + 18 = 36)
Total storage = 36tb

(4x 18 = 72)
(72 - 18 = 54)
(Par drive = 18 (18 + 18 +18 = 54)
Total storage = 54tb

Extending pool
(3x 18 = 54)
(54 - 18 = 36)
(Par drive = 18 (18 + 18 = 36)
Total storage = 72tb (6x 18 HDD with 2x 18 par)

(4x 18 = 72)
(72 - 18 = 54)
(Par drive = 18 (18 + 18 + 18 = 54)
Total storage = 108tb (8x 18 HDD)

 

[Redcoat]

Have I got the calculations, correct?

Welcome to the forums!!

From our Resource Section: https://www.truenas.com/community/resources/openzfs-capacity-calculator.185/

As well as the linked primary resource, take a look at the wintelguy link, too, for added background.

Good luck with your new server!

 

[Arwen]

Welcome as well.

Take a look at the ZFS terminology guide. The term "ZFS1" is a bit meaningless, though we can guess you mean RAID-Z1.

Terminology and Abbreviations Primer

Next, in general, RAID-Z1 is not recommended for larger disks, (basically over 1TB or 2TB). This has to do with another disk having a failure, (block or full disk), during bad disk replacement. So, 4 disks in a RAID-Z2 is somewhat the minimum. That said, if you have good backups or understand & accept the risks, RAID-Z1 with 18TB disks can work for you.

One thing ZFS does differently, is if redundancy for a file is lost because of a bad disk, and bad blocks on another disk, ZFS will tell you which files or directories are impacted. Thus, after you fix the hardware fault, you need only restore those file(s) to your pool. No need for a full restore of data. This means even with RAID-Z1 you may not loose your pool on 2 disk problems, potentially just a file or so.

ZFS does something else odd, compared to other RAID schemes. You can replace in place for disks that have not failed completely. Basically add the new, replacement disk to your server, and then perform the ZFS replacement. This allows any good data from the disk being replaced to be used. Even if a 2nd disk in your pool has a bad block. This can prevent the need to restore a file from backups.

Another thing that ZFS does that other RAID and file systems generally don't do, is redundant metadata. My media server uses a striped pool, (no data redundancy). So when I loose a media file, I have to restore it from backups. No real problem for me. However, one day I found that pool had a bad block, but no file was called out. Turns out I lost one of two metadata copies. So ZFS was able to use the second copy to re-create the first. No restore from backups was needed.

Directory entries & associated non-file information is part of metadata. Metadata is considered more important that file data because on loss of metadata that describes the file data, you can loose the file data. So, that is the reason why ZFS normally has at least 2 copies of metadata. (Critical metadata has 3 copies by default.) Plus, in general, metadata takes up much less space on disks that file data. These copies are on TOP of any vDev redundancy, like RAID-Z1/2/3.

 

[EFW]

Welcome to the forums!!

From our Resource Section: https://www.truenas.com/community/resources/openzfs-capacity-calculator.185/

As well as the linked primary resource, take a look at the wintelguy link, too, for added background.

Good luck with your new server!

Thanks, I checked that website.

Welcome as well.

Take a look at the ZFS terminology guide. The term "ZFS1" is a bit meaningless, though we can guess you mean RAID-Z1.

Yes, I did, thanks

Terminology and Abbreviations Primer

Next, in general, RAID-Z1 is not recommended for larger disks, (basically over 1TB or 2TB). This has to do with another disk having a failure, (block or full disk), during bad disk replacement. So, 4 disks in a RAID-Z2 is somewhat the minimum. That said, if you have good backups or understand & accept the risks, RAID-Z1 with 18TB disks can work for you.

One thing ZFS does differently, is if redundancy for a file is lost because of a bad disk, and bad blocks on another disk, ZFS will tell you which files or directories are impacted. Thus, after you fix the hardware fault, you need only restore those file(s) to your pool. No need for a full restore of data. This means even with RAID-Z1 you may not loose your pool on 2 disk problems, potentially just a file or so.

ZFS does something else odd, compared to other RAID schemes. You can replace in place for disks that have not failed completely. Basically add the new, replacement disk to your server, and then perform the ZFS replacement. This allows any good data from the disk being replaced to be used. Even if a 2nd disk in your pool has a bad block. This can prevent the need to restore a file from backups.

Another thing that ZFS does that other RAID and file systems generally don't do, is redundant metadata. My media server uses a striped pool, (no data redundancy). So when I loose a media file, I have to restore it from backups. No real problem for me. However, one day I found that pool had a bad block, but no file was called out. Turns out I lost one of two metadata copies. So ZFS was able to use the second copy to re-create the first. No restore from backups was needed.

Directory entries & associated non-file information is part of metadata. Metadata is considered more important that file data because on loss of metadata that describes the file data, you can loose the file data. So, that is the reason why ZFS normally has at least 2 copies of metadata. (Critical metadata has 3 copies by default.) Plus, in general, metadata takes up much less space on disks that file data. These copies are on TOP of any vDev redundancy, like RAID-Z1/2/3.

That all sounds great, thanks for the info. Once my trueNAS box is up and running, will be running a backup server which will take nightly snapshots and store them on the backup server, so I believe I will be ok with RAID-Z1.