Do not interrupt ShardingDDLCoordinator instances before releasing distributed locks

XMLWordPrintableJSON

    • Fully Compatible
    • ALL
    • Sharding EMEA 2021-05-17
    • None
    • 3
    • None
    • None
    • None
    • None
    • None
    • None
    • None

      Failed sharding DDL coordinators that didn't managed to write their state document on disk are explicitly de-registered from the in-memory registry by calling POS::releaseInstance().
      The problem is that if we pass a failed status to this function we will trigger the interruption of the instance before we managed to release the distributed locks previously acquired from that instance.

              Assignee:
              Tommaso Tocci
              Reporter:
              Tommaso Tocci
              Votes:
              0 Vote for this issue
              Watchers:
              2 Start watching this issue

                Created:
                Updated:
                Resolved: