Uploaded image for project: 'Core Server'
  1. Core Server
  2. SERVER-39826

Add a new index build failpoint to leave indexes unfinished and replace uses of hangAfterStartingIndexBuildUnlocked missing an 'off' call

    • Type: Icon: Task Task
    • Resolution: Fixed
    • Priority: Icon: Major - P3 Major - P3
    • 4.1.9
    • Affects Version/s: None
    • Component/s: Storage
    • None
    • Fully Compatible
    • Storage NYC 2019-02-25, Storage NYC 2019-03-11

      hangAfterStartingIndexBuildUnlocked is currently used to hang an index build on a secondary until shutdown in a couple JS tests. This works when run on a [repl-index-builder-X] thread because the process exits on shutdown and those threads stop being scheduled. The IndexBuildsCoordinator, however, will wait for threads to finish on shutdown, so this behavior will hang shutdown once we move builds onto the Coordinator.

      So instead of hanging until shutdown, we will add a failpoint that will cleanly interrupt the index build without cleaning up its persisted memory, just like clean shutdown does normally when an active index build throws InterruptedAtShutdown.

            Assignee:
            dianna.hohensee@mongodb.com Dianna Hohensee (Inactive)
            Reporter:
            dianna.hohensee@mongodb.com Dianna Hohensee (Inactive)
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

              Created:
              Updated:
              Resolved: