Uploaded image for project: 'Couchbase Server'
  1. Couchbase Server
  2. MB-7017

XDCR: extremely high number of connections in bidirectional 4->4 nodes performance tests

    XMLWordPrintable

Details

    • Bug
    • Resolution: Fixed
    • Blocker
    • 2.0
    • 2.0-beta-2
    • couchbase-bucket
    • Security Level: Public
    • build 1885 vs. build 1893
      VMs, CentOS 5.8, bidir
      4-to-4 nodes; 3-to-3 nodes with rebalance to 3-to-4 nodes.

    Description

      During the test number of connections to the server is continuously growing.
      Even after performance test when clusters are idle number of connection is very high. Also it varies across different server: from 40 to 1000.

      Diags for rebalance case:
      https://docs.google.com/file/d/0BwLUessE73UXOEJaRl9oSHVSbDQ/edit

      Attachments

        1. 10.2.2.218.7z
          18.43 MB
          Pavel Paulau
        2. 10.2.2.219.7z
          19.05 MB
          Pavel Paulau
        3. 10.2.2.220.7z
          18.47 MB
          Pavel Paulau
        4. 10.2.2.221.7z
          18.65 MB
          Pavel Paulau
        5. all.log
          190 kB
          Pavel Paulau
        6. diag.7z
          18.27 MB
          Pavel Paulau
        7. ensure_full_commit.log
          731 kB
          Junyi Xie
        8. memcached.log
          53 kB
          Pavel Paulau
        9. out
          725 kB
          Aleksey Kondratenko
        10. output
          77 kB
          Aleksey Kondratenko
        11. rebalance_test.png
          36 kB
          Pavel Paulau
        12. regular_bidir_test.png
          37 kB
          Pavel Paulau
        13. Screen Shot 2012-10-30 at 2.24.32 PM.png
          335 kB
          Junyi Xie
        14. script.rb
          0.5 kB
          Aleksey Kondratenko
        For Gerrit Dashboard: MB-7017
        # Subject Branch Project Status CR V
        21984,2 MB-7017: close tcp connection after persisting priority checkpoint master ns_server Status: ABANDONED -1 +1
        22105,3 MB-7017 Separate entry per high priority checkpoint flush request master ep-engine Status: MERGED +2 +1
        22191,2 MB-7017 Reduce the max timeout of checkpoint persistence to 30 sec master ep-engine Status: MERGED +2 +1

        Activity

          People

            chiyoung Chiyoung Seo (Inactive)
            pavelpaulau Pavel Paulau (Inactive)
            Votes:
            0 Vote for this issue
            Watchers:
            4 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved:

              Gerrit Reviews

                There are no open Gerrit changes

                PagerDuty