Issue Summary

Flush Edge Index Queue job runs into an error. At some point, this causes an instance to be out of memory.

Steps to Reproduce

1. User A creates a page
2. User B likes the page
3. User A changes page permission - view and edit for User A only
4. Wait for the Flush Edge Index Queue job to run
5. Error can be seen in the logs but it need to run for a long time for GC to spike

Expected Results

The error should be thrown and it should not cause an instance to be out OOM

Actual Results

2022-07-15 11:03:47,472 ERROR [Caesium-1-1] [impl.schedule.caesium.JobRunnerWrapper] runJob Scheduled job com.atlassian.confluence.plugins.confluence-edge-index:flushEdgeIndexQueueJob#flushEdgeIndexQueue failed to run
com.atlassian.bonnie.LuceneException: org.springframework.transaction.UnexpectedRollbackException: Transaction rolled back because it has been marked as rollback-only

As soon as the first entry is recorded, the heap utilization increases.

Workaround

Rebuild index from scratch but the issue can happen again after some time

Note:

Disabling the Flush Edge Index Queue job stops the error in the logs but heap utilization doesn't drop