MigrationBlockingOperationCoordinator may leave an inconsistent catalog between retries

XMLWordPrintableJSON

    • Cluster Scalability
    • Fully Compatible
    • ALL
    • v9.0
    • ClusterScalability 6Jul-20Jul, ClusterScalability 20Jul-3Aug
    • 200
    • None
    • None
    • None
    • None
    • None
    • None
    • None

      MigrationBlockingOperationCoordinator is effectively managed externally by the MultiUpdateCoordinator. When MigrationBlockingOperationCoordinator calls allowMigrations and fails, it simply bails out and the coordinator is torn down. The "parent" MultiUpdateCoordinator takes care of retrying the coordinator (by creating a new one), but since the old coordinator was destroyed, it released the DDL locks, which means that if setAllowChunkOperations left the global and shard catalogs in an inconsistent state, that inconsistency is visible to CMC. MigrationBlockingOperationCoordinator should make sure to call allowMigrations until it returns success without releasing the DDL lock.

            Assignee:
            Brett Nawrocki
            Reporter:
            Aitor Esteve Alvarado
            Votes:
            0 Vote for this issue
            Watchers:
            4 Start watching this issue

              Created:
              Updated:
              Resolved: