> On Sep 8, 2017, at 10:39 AM, Dan Langille <d...@langille.org> wrote:
>
>> On Sep 7, 2017, at 11:45 PM, Phil Stracchino <ph...@caerllewys.net
>> <mailto:ph...@caerllewys.net>> wrote:
>>
>> On 09/07/17 22:04, Dan Langille wrote:
>>> I have recently moved to one Storage device per pool.
>>>
>>> Why?
>>>
>>> So each pool is on a different ZFS dataset.
>>
>>
>> I'm curious. What do you see as the advantage of this layout?
>
>
> Background: I'm using ZFS on FreeBSD. I like keeping application specific
> data
> in its own ZFS dataset. It allows for an easy snapshot and backup solution.
>
> Storage space was the original motivation. One zpool was at 80%,
> the other is now at about 44%. I wanted to move some data from one zpool
> to another.
>
> When moving to another pool, the mount point changes, which meant creating
> a new Storage device pointing at that mount point.
>
> After moving the first Pool I figured: well, if it's this easy, why not do it
> for each Bacula Pool?
>
> Back to your question: A similar question was asked on Google+:
>
> https://plus.google.com/+DanLangille/posts/K2zfNo1ii22
> <https://plus.google.com/+DanLangille/posts/K2zfNo1ii22>
>
> ZFS can certainly and easily handle many files in the same directory. That
> wasn't
> a basis for change.
>
> My answer from there (with slight edits) appears below:
>
> ###
>
> If the Pool is no longer required, it is easy to delete. Dividing data up
> this way
> it's always a good idea, because of the flexibility for future manipulation
> it provides.
>
> If these were all on separate devices, you'd get better concurrent throughput.
>
> Full backups are usually bigger, incremental backups are usually smaller, so
> you
> could just recordsize accordingly.
>
> If you assigned each client to a different pool, deleting their backups when
> the
> client leaves is now a simple matter of deleting the appropriate ZFS datasets.
>
> If you want to move a pool to a different bacula-sd, you move that data set.
>
> ###
>
> The above was typed very early in my day and I wasn't quite awake.
>
> Perhaps the highlights are:
>
> - keep like data in like datasets
> - more datasets gives more flexibility
> - because I can (I was at vBSDCon and had the time to do this)
>
> Ideas? Suggestions?
Over a discussion last night at dinner, we came up with two additional benefits:
* snapshot retention - If you want a different snap shot retention policy
for different types of backups, having each pool on a different ZFS dataset
allows this.
* zfs replication - if you want to replicate some, but not all of your backups,
having only that
data on a separate ZFS dataset allows for this.
--
Dan Langille - BSDCan / PGCon
d...@langille.org
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users
