When a node tries to register a cluster lock, it will try to register it in a local memory, and then register it in the database, which we expect may fail as another node may have already created this entry.
The problem happens if this code (that is creating the cluster lock) is executed within an existing database transaction. In that case, the whole transaction is marked to be rolled back, rather than accepting the error as part of the algorithm.
- Jira Data Center 8.3+
This issue can be reproduced with regular actions such as:
- archiving multiple issues in a row.
- transitioning issues.
Operations would complete with no errors.
Users may see an error on the screen:
These exceptions are thrown in the atlassian-jira.log file:
Currently there is no known workaround for this behavior. A workaround will be added here when available