[talk] zfs disk outage and aftermath

John C. Vernaleo john at netpurgatory.com
Mon Feb 4 16:56:59 EST 2019


I read the subject of this had that sinking feeling in my stomach as I'm 
becoming more and more reliant on ZFS and no email with 'disk' in the 
subject line is ever good news.  I was afraid there would be a horror 
story to make me rethink my reliance on ZFS,

Was pleasantly surprissed to see I was wrong.  And that health check 
script looks like a really good idea.

John

-------------------------------------------------------
John C. Vernaleo, Ph.D.
www.netpurgatory.com
john at netpurgatory.com
-------------------------------------------------------

On Mon, 4 Feb 2019, N.J. Thomas wrote:

> Well, yesterday morning I suffered my first drive failure on one of my
> ZFS boxes (running FreeBSD 12.0), it actually happened on my primary
> backup server.
>
> "zpool status" showed that my regularly scheduled scrub had found (and
> fixed) some errors on one of the disks in a mirrored pair.
>
> I made sure that my replica box had up to date snapshots transferred
> over, shut down the machine, and asked the datacenter team to check.
> They indeed found that the drive was faulty and replaced it.
>
> It took about 4 hours for the drive to be resilvered, and that was it.
> Back to normal with almost no issues -- apart from the few minutes that
> the machine was down while its drive was being replaced.
>
> My takeaways:
>
>    - use ZFS
>
>    - take regular snapshots
>
>    - replicate your snapshots to another machine
>
>    - scrub your disks regularly (unlike fsck, this can be run while the
>      drive is mounted and active)
>
>    - monitor zfs health (I use this script from Calomel.org:
>      https://calomel.org/zfs_health_check_script.html)
>
> The first three points are kinda obvious, the last two I picked up from
> other, more experienced, ZFS users.
>
> I had been waiting for this day since I first started using ZFS years
> ago and am very happy with that decision to use this filesystem.
>
> Thomas
>
> _______________________________________________
> talk mailing list
> talk at lists.nycbug.org
> http://lists.nycbug.org:8080/mailman/listinfo/talk
>
>



More information about the talk mailing list