2:51 a.m.
Hello,
I found that in the LUG2015 talk by Gabriele Paciucci states that:
"Resilvering and scrubbing are autotuned by the ZFS I/O scheduler"
When I trigger
lctl lfsck_start -M zaip-MDT0000 --all
the zpool status does not show the scrub progress.
As I understood from the Lustre Manual the lfsck_start fixes the lustre
relevant part but not zfs,
Should I schedule the zpool scrub in order to avoid blocks corruption?
Thanks,
Arman.
***********************************************************
Dr. Arman Khalatyan eScience -SuperComputing
Leibniz-Institut für Astrophysik Potsdam (AIP)
An der Sternwarte 16, 14482 Potsdam, Germany
***********************************************************
4:30 a.m.
On 2016/05/31, 03:51, "Arman Khalatyan" <arm2arm(a)gmail.com> wrote:
Hello,
I found that in the LUG2015 talk by Gabriele Paciucci states that:
"Resilvering and scrubbing are autotuned by the ZFS I/O scheduler"
When I trigger
lctl lfsck_start -M zaip-MDT0000 --all
the zpool status does not show the scrub progress.
As I understood from the Lustre Manual the lfsck_start fixes the lustre
relevant part but not zfs,
Should I schedule the zpool scrub in order to avoid blocks corruption?
Yes. This is the same as regular ZFS. It needs to be scrubbed periodically to detect and
correct any latent errors in the storage.
At one point there was a discussion about integrating the LFSCK iteration
with the ZFS scrub to reduce the amount of background IO to the storage,
but this has not been implemented.
Cheers, Andreas
--
Andreas Dilger
Lustre Principal Architect
Intel High Performance Data Division
5:47 a.m.
Thank you!
***********************************************************
Dr. Arman Khalatyan eScience -SuperComputing
Leibniz-Institut für Astrophysik Potsdam (AIP)
An der Sternwarte 16, 14482 Potsdam, Germany
***********************************************************
On Tue, May 31, 2016 at 1:30 PM, Dilger, Andreas <andreas.dilger(a)intel.com>
wrote:
On 2016/05/31, 03:51, "Arman Khalatyan"
<arm2arm(a)gmail.com> wrote:
> Hello,
> I found that in the LUG2015 talk by Gabriele Paciucci states that:
> "Resilvering and scrubbing are autotuned by the ZFS I/O scheduler"
> When I trigger
>
> lctl lfsck_start -M zaip-MDT0000 --all
>
> the zpool status does not show the scrub progress.
> As I understood from the Lustre Manual the lfsck_start fixes the lustre
> relevant part but not zfs,
> Should I schedule the zpool scrub in order to avoid blocks corruption?
Yes. This is the same as regular ZFS. It needs to be scrubbed
periodically to detect and correct any latent errors in the storage.
At one point there was a discussion about integrating the LFSCK iteration
with the ZFS scrub to reduce the amount of background IO to the storage,
but this has not been implemented.
Cheers, Andreas
--
Andreas Dilger
Lustre Principal Architect
Intel High Performance Data Division
2187
days inactive
2187
days old
2 comments
2 participants
participants (2)
-
Arman Khalatyan -
Dilger, Andreas