Uploaded image for project: 'Couchbase Server'
  1. Couchbase Server
  2. MB-30234

Major throughput degradation in data insertion scenarions

    XMLWordPrintable

Details

    • Bug
    • Resolution: Fixed
    • Critical
    • 5.5.0
    • 5.5.0
    • couchbase-bucket
    • reproduced across various tests and clusters.

    Description

      The issue has been introduced in build 2886 by following change:

      {{kv_engine changed from 1aae4b168c2ab28d0f9657117280aed5be0acfd3 to vulcan }}
      [+] 075614a MB-30019: Always close replica-checkpoint on memory-snapshot-end

       

      Behavior:

      The DCP (replication) backlog grows very quickly and then at some point replication rate drops dramatically. Probably because it starts reading data from disk rather that memory. Here is good representation of the issue:

       

      DCP backlog:

       

       

      Sets per sec:

       

       

       

      Disk utilization:

       

       

       

       

      Attachments

        1. dcp_backlog.png
          dcp_backlog.png
          64 kB
        2. disk_utilization.png
          disk_utilization.png
          103 kB
        3. set_ops_per_sec.png
          set_ops_per_sec.png
          57 kB

        Issue Links

          relates to

          Bug - A problem which impairs or prevents the functions of the product. MB-30019 High vb_replica_checkpoint_memory at the end of data ingestion

          • Major - Major loss of function.
          • Closed

          Activity

            People

              oleksandr.gyryk Alex Gyryk (Inactive)
              oleksandr.gyryk Alex Gyryk (Inactive)
              Votes:
              0 Vote for this issue
              Watchers:
              4 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved:

                PagerDuty