Uploaded image for project: 'Core Server'
  1. Core Server
  2. SERVER-30841

Lower the amount of metadata refresh logging

    • Type: Icon: Improvement Improvement
    • Resolution: Fixed
    • Priority: Icon: Major - P3 Major - P3
    • 3.6.8, 4.0.3, 4.1.1
    • Affects Version/s: 3.4.0, 3.5.1, 3.6.3
    • Component/s: Sharding
    • Labels:
    • Fully Compatible
    • v4.0, v3.6
    • Sharding 2018-04-23, Sharding 2018-05-07, Sharding 2018-07-30

      Example of chatty logging:

      2017-08-22T12:02:14.386+1000 I SHARDING [Balancer] Refreshing chunks for collection mytest.test1 based on version 0|0||000000000000000000000000
      2017-08-22T12:02:14.387+1000 I SHARDING [CatalogCacheLoader-0] Refresh for collection mytest.test1 took 0 ms and found version 1|0||599b909d11c6b22dd4ad5548

      These logs don't convey knowledge to our users and are not actionable. On the config server, they can spam to a user with many sharded collection in our auto-balancing code every 10 seconds because we force refresh for every chunk the balancer attempts to move.

      The logs come from the CatalogCache and the CatalogCacheLoader.

      The logs were useful while we were trying to stabilize the new/modified refresh logic, but less so now, and our testing can retain the same information with log-level 1 set.

            cheahuychou.mao@mongodb.com Cheahuychou Mao
            dianna.hohensee@mongodb.com Dianna Hohensee (Inactive)
            6 Vote for this issue
            14 Start watching this issue