Uploaded image for project: 'Core Server'
  1. Core Server
  2. SERVER-89875

Consider stopping the checkpoint thread prior to quiesce mode.

    • Type: Icon: Improvement Improvement
    • Resolution: Unresolved
    • Priority: Icon: Major - P3 Major - P3
    • None
    • Affects Version/s: None
    • Component/s: None
    • None
    • Storage Execution

      This is an idea I had from a help ticket which I will link to this ticket. Essentially there was a very long shutdown duration. I noted that while the server was in quiesce mode, regular WiredTiger checkpoints were still being created. The quiesce mode would then finish but the server would be waiting for the normal WT checkpoint to finish (which took an extra 2 minutes in the example) before finally closing WiredTiger which then triggers a final shutdown checkpoint.

      If we prevented new checkpoints from starting during the quiesce phase we could at least make this shutdown flow more deterministic. I don't think that this would necessarily be faster as some of the work not done by the checkpoint during quiesce would then be done on shutdown.

      Anyway this was an idea and this project seemed like the best place to put it. Feel free to close it if it doesn't like a beneficial idea.

            Assignee:
            Unassigned Unassigned
            Reporter:
            luke.pearson@mongodb.com Luke Pearson
            Votes:
            0 Vote for this issue
            Watchers:
            11 Start watching this issue

              Created:
              Updated: