Hello folks. I’ve recently upgraded a cluster from Nautilus to Pacific (16.2.7), and ever since then I’ve noticed that scrubs don’t seem to make progress. Even though there are hundreds of PGs in "scrubbing+deep” state, the humber of "pgs not deep-scrubbed in time” has been steadily increasing, which did not happen before the upgrade. Also, PGs appear to be many hours in this state, and the disk I/O is practically non-existent. Also, since the upgrade OSDs have been spamming the logs with messages like "handle_scrub_reserve_grant: received unsolicited reservation grant”, could this be related? I’ve tried stopping all scrubs and deep-scrubs (even had to restart some OSDs to force the scrub to stop) and these messages went away, but after unsetting noscrub and nodeep-scrub the problem came back. The only reference I’ve seen to this issue was https://www.mail-archive.com/ceph-users@xxxxxxx/msg13997.html <https://www.mail-archive.com/ceph-users@xxxxxxx/msg13997.html>, but it mentions some fix that hasn’t been back ported yet. Is there a status on this? All the OSDs still report legacy omap stats "OSD(s) reporting legacy (not per-pool) BlueStore omap usage stats”, I haven’t yet restarted them with the flag to fix at startup. Could this somehow also be related? Would appreciate any pointers, thank you! André _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |