Description
The way we currently lock archives in cbbackupmgr can keep stale lockfiles around if cbbackupmgr fails to sync with the remote archive (e.g. CBSE-14817), preventing other instances of cbbackupmgr from operating on that archive if they have a different host name. There may be better ways we can lock archives which reduce the probability of stale lockfiles being left around.
Considerations
- The "lock" should still work. Multiple cbbackupmgr instances running on the same archive would be very bad. Better to have stale lockfiles which need to be manually deleted than a broken repo/archive.
- Any improvements should be backwards compatible i.e. would this still work if there was another backup node in the cluster running an older version?
Attachments
Issue Links
- relates to
-
MB-53626 [CBM] Log errors when trying to unlock object storage lock file
- Closed