Uploaded image for project: 'Couchbase Server'
  1. Couchbase Server
  2. MB-19812

XDCR has replication throughput regression in Spock

    XMLWordPrintable

Details

    • Untriaged
    • Centos 64-bit
    • Unknown

    Description

      Running a unidirectional XDCR test with no load we see that the replication rate in 4.5 and prior release is around 60,000 items/sec and in Spock it is 2500 items/sec.

      Logs are attached. Replication is from 10.17.0.105 to 10.17.0.106.

      Please let me know if there is more information I can provide.

      Attachments

        No reviews matched the request. Check your Options in the drop-down menu of this sections header.

        Activity

          This was seen on 4.7.0-711, better performance is seen on all Watson builds.

          Which DCP data should I get?

          ericcooper Eric Cooper (Inactive) added a comment - This was seen on 4.7.0-711, better performance is seen on all Watson builds. Which DCP data should I get?

          xdcr log files showed that XMEM was the bottleneck. DCP was fine.

          yu Yu Sui (Inactive) added a comment - xdcr log files showed that XMEM was the bottleneck. DCP was fine.
          jliang John Liang added a comment -

          We run perf test for 100M bucket on both regular xdcr replication and optimistic replication. In both cases, we see a drop in throughput, even though there is no code difference from xdcr. Here are the 2 xdcr runs:

          http://perf.jenkins.couchbase.com/job/xdcr-5x5/2647/
          http://perf.jenkins.couchbase.com/job/xdcr-5x5/2646/

          In addition, there are open spock performance bug on KV

          https://issues.couchbase.com/browse/MB-19821

          jliang John Liang added a comment - We run perf test for 100M bucket on both regular xdcr replication and optimistic replication. In both cases, we see a drop in throughput, even though there is no code difference from xdcr. Here are the 2 xdcr runs: http://perf.jenkins.couchbase.com/job/xdcr-5x5/2647/ http://perf.jenkins.couchbase.com/job/xdcr-5x5/2646/ In addition, there are open spock performance bug on KV https://issues.couchbase.com/browse/MB-19821
          jliang John Liang added a comment -

          I am resolving the bug for now as dup. From xdcr logs, we also have not found any issue. Given that there is no change in xdcr code, i don't expect that should be a big regression between watson and spock. Pls re-open the bug if you still see regression after KV issue are fixed.

          jliang John Liang added a comment - I am resolving the bug for now as dup. From xdcr logs, we also have not found any issue. Given that there is no change in xdcr code, i don't expect that should be a big regression between watson and spock. Pls re-open the bug if you still see regression after KV issue are fixed.

          This is a bulk update to close all invalid, won't-fix and duplicate bugs. If you think your bug was closed incorrectly, please feel free to reopen

          raju Raju Suravarjjala added a comment - This is a bulk update to close all invalid, won't-fix and duplicate bugs. If you think your bug was closed incorrectly, please feel free to reopen

          People

            yu Yu Sui (Inactive)
            ericcooper Eric Cooper (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