Uploaded image for project: 'Core Server'
  1. Core Server
  2. SERVER-22642

WiredTiger engine resync stalls with a lot of tables/indexes

    • Type: Icon: Bug Bug
    • Resolution: Duplicate
    • Priority: Icon: Major - P3 Major - P3
    • None
    • Affects Version/s: 3.0.9, 3.2.1
    • Component/s: WiredTiger
    • Labels:
    • ALL

      We are trying to upgrade a replica set with many collections and indexes to WiredTiger. The replica set has 30k collections and 14 indexes per collection.

      The initial data sync and index build works fine but the server makes very little progress once it starts to apply the replication log. It seems to make no progress for ~15 minutes during this time it consumes a full core of CPU and does very little I/O. It will does a burst of I/O for a minute or so before falling back to consuming a lot of CPU.

      Based on the attached perf profile it looks like most of the CPU is being consumed by the eviction thread. (Possibly during a checkpoint?)

      Let me know if there is any additional information I can provide to help track this down.

        1. diagnostic.data_3.2.4-105-g73290d0.tar.bz2
          23.39 MB
          Bob Potter
        2. diagnostic.data.tar.gz
          21.46 MB
          Bob Potter
        3. mongo_32_perf.txt
          562 kB
          Bob Potter
        4. perf_report_3.2.4-105-g73290d0.txt.gz
          687 kB
          Bob Potter

            kelsey.schubert@mongodb.com Kelsey Schubert
            bpot Bob Potter
            0 Vote for this issue
            11 Start watching this issue