MigrationDestinationManager::abort() should interrupt blocked recipient

XMLWordPrintableJSON

    • Type: Bug
    • Resolution: Unresolved
    • Priority: Major - P3
    • None
    • Affects Version/s: None
    • Component/s: None
    • None
    • Cluster Scalability
    • ALL
    • 200
    • None
    • None
    • None
    • None
    • None
    • None
    • None

      When a migration recipient receives _recvChunkAbort while blocked waiting for orphan cleanup, MigrationDestinationManager::abort() does not interrupt the operation. The migration thread continues waiting for the full receiveChunkWaitForRangeDeleterTimeoutMS (up to 300s), holding ActiveMigrationsRegistry and stalling any operation waiting for migrations to drain.

      This was latent until SERVER-114851, which started pausing orphan deletions during resharding. More details on how this bug manifested.

            Assignee:
            Kruti Shah
            Reporter:
            Kruti Shah
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

              Created:
              Updated: