.Backtrace now works as expected for CephFS scrub operations
Previously, backtrace was unwritten to stable storage. Scrub activity reported a failure if the backtrace did not match the in-memory copy for a new and unsynced entry.
Backtrace mismatch also happened for a stray entry that was about to be purged permanently since there was no need to save the backtrace to the disk. Due to the ongoing metadata I/O, it might have happened that the raw stats would not match if there was heavy metadata I/O because the raw stats accounting is not instantaneous.
To workaround this issue, rerun the scrub when the system is idle and has had enough time to flush in-memory state to disk. As a result, once the metadata has been flushed to the disk, these errors are resolved. Backtrace validation is successful if there is no backtrace found on the disk and the file is new, and the entry is stray and about to be purged.
See the KCS link:https://access.redhat.com/solutions/6123271[_Ceph status shows HEALTH_ERR with MDSs report damaged metadata_] for more details.
Comment 54Patrick Donnelly
2021-06-16 16:13:35 UTC
(In reply to Michael J. Kidd from comment #52)
> I've created a KCS with the relevant diagnostic and repair steps:
> https://access.redhat.com/solutions/6123271
>
> Please let me know if there are any changes needed.
Looks good! Thanks Michael.
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory (Red Hat Ceph Storage 5.0 bug fix and enhancement), and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.
https://access.redhat.com/errata/RHBA-2021:3294