Uploaded image for project: 'Core Server'
  1. Core Server
  2. SERVER-45400

Reduce OplogFetcher getMore awaitData timeout for nodes in RollbackTest fixture

    • Type: Icon: Improvement Improvement
    • Resolution: Fixed
    • Priority: Icon: Major - P3 Major - P3
    • 4.3.3
    • Affects Version/s: None
    • Component/s: Replication
    • Labels:
    • Fully Compatible
    • v4.2, v4.0
    • Repl 2020-01-27

      SERVER-37390 made it so that the RollbackTest fixture uses a high election timeout (24 hours) to prevent unplanned state transitions during test execution. This has the unintended side effect of slowing down tests that use this fixture since the awaitData timeout on getMores sent by the OplogFetcher are affected by the election timeout. A 24 hour election timeout causes us to use the maximum awaitData timeout which is 30 seconds. When we stop replication throughout the RollbackTest, this can cause delays when we wait for the stopReplProducer failpoint to be hit, since a getMore doesn't time out and return to the syncing node for 30 seconds even if there is no data to fetch. We can reduce this timeout inside RollbackTest by using this existing failpoint.

            ryan.timmons@mongodb.com Ryan Timmons
            william.schultz@mongodb.com William Schultz
            0 Vote for this issue
            4 Start watching this issue