Uploaded image for project: 'Core Server'
  1. Core Server
  2. SERVER-38310

Oplog information collection stalls FTDC

    XMLWordPrintable

Details

    • Bug
    • Status: Closed
    • Major - P3
    • Resolution: Fixed
    • None
    • 4.1.7
    • None
    • None
    • Fully Compatible
    • ALL
    • Security 2018-12-17

    Description

      In SERVER-37625 we added oplog information collection to ftdc in order to have oplog window information (oplog earliest and latest times). This has an unfortunate side effect of stalling ftdc when the system is under stress in waitForAllEarlierOplogWritesToBeVisible while reading the earliest and latest documents from the oplog. This is bad because when the system is under stress is exactly when we need ftdc the most.

      Generally the oplog is large enough that earliest and latest times don't change very quickly relative to the size of the oplog, so it would be ok if this information was a little stale. Maybe we can use a (short) timeout on this read from the oplog, and if it times out use the last collected information? Or maybe we can do this read without calling waitForAllEarlierOplogWritesToBeVisible because for statistics purposes (I think) it doesn't really matter whether earlier oplog writes are visible?

      Attachments

        Issue Links

          Activity

            People

              mark.benvenuto@mongodb.com Mark Benvenuto
              bruce.lucas@mongodb.com Bruce Lucas
              Votes:
              0 Vote for this issue
              Watchers:
              12 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved: