Ceph » RADOS

/a/dzafman-2018-03-21_09:57:19-rados:thrash-wip-zafman-testing2-distro-basic-smithi/2312125

rados:thrash/{0-size-min-size-overrides/3-size-2-min-size.yaml 1-pg-log-overrides/short_pg_log.yaml backoff/peering_and_degraded.yaml ceph.yaml clusters/{fixed-2.yaml openstack.yaml} d-balancer/upmap.yaml msgr-failures/osd-delay.yaml msgr/simple.yaml objectstore/filestore-xfs.yaml rados.yaml rocksdb.yaml thrashers/morepggrow.yaml thrashosds-health.yaml workloads/small-objects.yaml}

Just bytes

dzafman-2018-03-28_18:21:29-rados-wip-zafman-testing-distro-basic-smithi/2332093

[ERR] 3.0 scrub stat mismatch, got
51/51 objects, 4/4 clones, 24/24 dirty, 0/0 omap, 0/0 pinned, 1/1 hit_set_archive,
7/7 whiteouts, 119951885/120679921 bytes, 0/0 manifest objects, 269/269 hit_set_archive
bytes

UPDATE: this is cache tiering and bytes, moved to #22656

The original report was an EC test, so it looks like a dup of #23339.

David, your failures are not EC. Could they be from your testing branch? I don't think I've seen them in master.

My change only affects the scrub error counts in the stats. However, if setting dirty_info in proc_primary_info() would cause wrong stats to be written along with the scrub change, then that could cause this problem. So this could be caused by my change, but I'm not sure how it would be fixed.

If this is because dirty_info is set it would mean this is a latent defect where if history or purge_snaps need to be updated then we could run into this too.