We just got a reply from Intel telling us that there's a new firmware
coming out soon to fix an issue where S4510 and S4610 drives get IO
timeouts that may lead to drive drops when under heavy load. This might
very well be the source of our issue.
On 9/23/22 11:12, Stefan Kooman wrote:
On 9/23/22 15:22, J-P Methot wrote:
Thank you for your reply,
discard is not enabled in our configuration as it is mainly the
default conf. Are you suggesting to enable it?
No. There is no consensus if enabling it is a good idea (depends on
proper implementation among other things). From my experience on Intel
S4610 (during LVM cleanup at OSD reprovisioning) it spends a lot of
time discarding blocks (IIRC an order of magnitude more than Samsung
PM883). So I doubt it would help. Altough it's hard to tell. Maybe it
can do discards more often it does take less per operation, and might
be less impactfull. But this is all speculation. It might be firmware
related. Have all disks the same firmware? Are there disks that never
experience this problem?
Gr. Stefan
--
Jean-Philippe Méthot
Senior Openstack system administrator
Administrateur système Openstack sénior
PlanetHoster inc.
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
