Uploaded image for project: 'WiredTiger'
  1. WiredTiger
  2. WT-2702

Under high thread load, WiredTiger exceeds cache size

    Details

    • Type: Bug
    • Status: Closed
    • Priority: Critical - P2
    • Resolution: Fixed
    • Affects Version/s: None
    • Fix Version/s: WT2.9.0, 3.2.8, 3.3.9
    • Labels:
      None

      Description

      • cache usage reaches 100% and continues climbing, reaching 145% by the end of the data
      • doesn't look like a long-running transaction (range of ids pinned doesn't grow)
      • no problem with amount of dirty data
        & checkpoints are proceeding at a regular pace, taking no more than about 10 s
      • at about the point where cache usage began to climb we see lag on secondary member 3 growing (but not on secondary member 2). This suggests maybe related to SERVER-23622, but when member 3 is restarted at D it is able to catch up quickly, and cache growth does not stop, actually appears to accelerate at that point. Not clear why it began lagging but was able to catch up quickly when it was restarted.

        Attachments

        1. WiredTigerStat.bad
          3.06 MB
        2. WiredTigerStat.patched
          4.20 MB
        3. wtperf-excess-cache.png
          wtperf-excess-cache.png
          171 kB
        4. wtperf-patched.png
          wtperf-patched.png
          171 kB

          Issue Links

            Activity

              People

              • Assignee:
                michael.cahill Michael Cahill
                Reporter:
                michael.cahill Michael Cahill
              • Votes:
                0 Vote for this issue
                Watchers:
                9 Start watching this issue

                Dates

                • Created:
                  Updated:
                  Resolved: