Details

    • Type: Bug
    • Status: Closed
    • Priority: Blocker - P1
    • Resolution: Fixed
    • Affects Version/s: 3.2.0-rc4
    • Fix Version/s: 3.2.0-rc5
    • Component/s: WiredTiger
    • Labels:
    • Backwards Compatibility:
      Fully Compatible
    • Operating System:
      ALL
    • Steps To Reproduce:
      Hide

      Standalone mongodb instance, 24CPU, 32GB RAM
      benchRun 32 false true 100000
      (see attached script)

      Show
      Standalone mongodb instance, 24CPU, 32GB RAM benchRun 32 false true 100000 (see attached script)

      Description

      With the resolution of SERVER-21652, things are definitely much improved. However, I still see some heavy stalls during what appears to be dirty writeback. Maybe checkpoints, but not time-aligned as I'd expect?

      1. benchRun
        2 kB
        Martin Bligh
      2. metrics.2015-11-30T14-44-15Z-00000
        120 kB
        Martin Bligh
      1. stalls.png
        154 kB
      2. stalls.png
        308 kB

        Issue Links

          Activity

          Hide
          pasette Dan Pasette added a comment -

          Are. You running with journal on separate volume?

          Show
          pasette Dan Pasette added a comment - Are. You running with journal on separate volume?
          Hide
          bruce.lucas Bruce Lucas added a comment -

          Also, FTDC data please? Want to check if it's similar to something Ramon showed me yesterday, involving a large number of slot join races during the slowdowns.

          Show
          bruce.lucas Bruce Lucas added a comment - Also, FTDC data please? Want to check if it's similar to something Ramon showed me yesterday, involving a large number of slot join races during the slowdowns.
          Hide
          martin.bligh Martin Bligh (Inactive) added a comment - - edited

          Dan Pasette Nope. fixing ... looking much better now. Do we still want to address that, or close this back out?

          Show
          martin.bligh Martin Bligh (Inactive) added a comment - - edited Dan Pasette Nope. fixing ... looking much better now. Do we still want to address that, or close this back out?
          Hide
          sue.loverso Sue LoVerso added a comment -

          Martin Bligh are you still seeing WT log stalls? Do you have data/stats from that? A large number of slot join races implies that we're waiting for a free slot. That means we may not be writing them out to the OS timely enough or we're stuck waiting for something.

          Show
          sue.loverso Sue LoVerso added a comment - Martin Bligh are you still seeing WT log stalls? Do you have data/stats from that? A large number of slot join races implies that we're waiting for a free slot. That means we may not be writing them out to the OS timely enough or we're stuck waiting for something.
          Hide
          martin.bligh Martin Bligh (Inactive) added a comment -

          Sue LoVerso Only if I put journal and data on the same device, otherwise it looks OK

          Show
          martin.bligh Martin Bligh (Inactive) added a comment - Sue LoVerso Only if I put journal and data on the same device, otherwise it looks OK

            People

            • Votes:
              0 Vote for this issue
              Watchers:
              16 Start watching this issue

              Dates

              • Created:
                Updated:
                Resolved: