Uploaded image for project: 'Core Server'
  1. Core Server
  2. SERVER-87307

ReshardingTxnCloner unnecessarily appends $incompleteOplogHistory oplog entry when retryability history is already marked incomplete

    • Cluster Scalability
    • Fully Compatible
    • ALL
    • Cluster Scalability 2024-4-1
    • 8
    • 3

      This causes the prevOpTime chain to grow and for reloading the prevOpTime chain from disk following a primary failover to take nominally longer. In extreme cases of repeated primary failovers seen in testing on slower build variants, this can prevent the ReshardingTxnCloner from making forward progress.

      We can skip calling resharding::data_copy::updateSessionRecord(TransactionParticipant::kDeadEndSentinel) when TransactionParticipant::hasIncompleteHistory() == true.

            Assignee:
            brett.nawrocki@mongodb.com Brett Nawrocki
            Reporter:
            max.hirschhorn@mongodb.com Max Hirschhorn
            Votes:
            0 Vote for this issue
            Watchers:
            5 Start watching this issue

              Created:
              Updated:
              Resolved: