Actions
Bug #18076
closedmultisite: some yields in RGWMetaSyncShardCR::full_sync() resume in incremental_sync()
% Done:
0%
Source:
Tags:
Backport:
jewel, kraken
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
When RGWMetaSyncShardCR::full_sync() is done, it updates the sync_marker and writes it to rados. But after it changes 'sync_marker.state = IncrementalSync;', any yields will cause RGWMetaSyncShardCR::operate() to call incremental_sync() instead - so most of the cleanup and error paths in full_sync() are unreachable.
Actions