[kwlug-disc] Filesystems for backups

[Mikalai Birukou]

On 2019-08-07 5:49 p.m., Paul Nijjar via kwlug-disc wrote:
> ZFS with FreeNAS or Ubuntu (using ZFS on Linux), raidz
> ------------------------------------------------------
>
> - Allows a fair amount of storage without full redundancy?

> - Does NOT allow me to add new drives easily. I can only add entire
>    sets that match the existing raidz array, turning RAID5 into RAID50?

May be I am reading your thoughts incorrectly, but I think that process 
here is not as strict.

a) You can always add a "device" to ZFS. It just adds a capacity. If any 
such higher level "device" is down, pool is down.

b) Device can be mirror, or raidz1-2-3, or may be just a disk.

Let's say today you have 5 disks, 1TB each. You decide to add these as a 
raidz1, i.e. 1 out of 5 disks out of these particular disks can be out 
without bringing down the whole higher level "device".

Next day you get 7 disks, 3TB each. You create a new device, say with 
raidz2, i.e. 2 of 7 disks out of the seven can fail without bringing 
down this second higher level "device". And we add new "device" to an 
existing pool.

Etc.

When disk fails in a "device", it has to be replaced with same or big 
size disk. When all disks in a device change to higher capacity, whole 
device can post greater capacity.

We may say that adding capacity of "devices" is what LVM does, i.e. 
creating a contiguous pool. In other words, for the purpose of 
assembling pools, ZFS = LVM + RAIDy at device level (modulo all great 
other features).

> - If a drive dies I can rebuild (with the usual RAID5 caveats)

Drive dies, detach+attach, and resilver? Note that rebuild actually 
happens inside of the "device".

Don't forget that ZFS allows you to scrub, checking data bytes, and 
finding errors that might've gone unnoticed, cause particular sectors 
with cold data are not touched as often. Its like cleaning teeth, very 
healthy crontab.


My opinion, ZFS it due to different types of disks.

If you wish, add SSD cache layer, or just keep ton of memory for ZFS 
cache to use. It will fly. By the way, addition of caching and memory 
adjustments can always be done later, while pool is online.