osd_scrub_sleep option blocks op thread in jewel + later
Just like snap trimming, when scrubbing was moved into the unified op queue, its sleep option was not changed, so it now blocks osd ops rather than just pausing scrubbing.
This sleep should be moved from PG::scrub() into the scrub state machine, and made async, just like it was for snap trimming - see http://tracker.ceph.com/issues/19241
#10 Updated by Nathan Cutler 8 months ago
@Brad - yes, when you open the backport PR, just mention both backport tracker issues in the commit message ;-)
I say "when you open" because I looked at the cherry pick of PR#14886 to jewel. . . and it's non-trivial. It conflicts, e.g., with https://github.com/ceph/ceph/commit/b7e77996586daa46da6be0bfbb8c2b93e203a83a and I don't know if that's a dependency.
Can I assign the backport trackers to you?