Uploaded image for project: 'Core Server'
  1. Core Server
  2. SERVER-91708

Resharding should consider queued index builds waiting for active number of index builds to be below the threshold

    • Cluster Scalability
    • ALL
    • 200

      Marking as blocked on SERVER-91965. After that has been resolved, this can be revisited.

      ------

      Details in BF-33268. 

      A started index build has a condvar wait to wait for the number of active index builds on the shard to be less than the threshold. The code where we begin registering the index build in that function looks like is here (after the condvar wait) which will call registerIndexBuild to add the build to the activeIndexBuilds map, which the resharding donor coordinator checks as a part of the assertion. If the index build is still waiting in the condvar wait, resharding will successfully pass checks since there are no active index builds on that particular collection. We should wait for any of those queued index builds to complete before resharding.

            Assignee:
            Unassigned Unassigned
            Reporter:
            jason.zhang@mongodb.com Jason Zhang (Inactive)
            Votes:
            0 Vote for this issue
            Watchers:
            8 Start watching this issue

              Created:
              Updated: