Uploaded image for project: 'Couchbase Server'
  1. Couchbase Server
  2. MB-36980

disable mutation processing after first abort from flusher

    XMLWordPrintable

    Details

    • Triage:
      Untriaged
    • Is this a Regression?:
      Unknown

      Description

      As seen in MB-36948, if mutation processing continues after first abort from flusher, it could lead to a hang. When TK processes an abort, it doesn't update the last flushed TS. It is possible for some vbuckets to have flushed successfully. This causes TK to generate a TS which flusher can never flush and leads to a hang.

        Attachments

          Issue Links

          relates to

          Bug - A problem which impairs or prevents the functions of the product. MB-43008 [System Test] : index building for a partitioned primary index is stuck

          • Critical - Crashes, loss of data, severe memory leak.
          • Closed
          For Gerrit Dashboard: MB-36980
          # Subject Branch Project Status CR V
          118321,1 MB-36980 Disable stream processing after first abort unstable indexing Status: ABANDONED 0 0
          141922,3 MB-36980 force recovery on flush abort unstable indexing Status: MERGED +2 +1

            Activity

            Ascending order - Click to sort in descending order
            Hide
            Permalink
            build-team Couchbase Build Team added a comment -

            Build couchbase-server-7.0.0-4090 contains indexing commit 48c0808 with commit message:
            MB-36980 force recovery on flush abort

            Show
            build-team Couchbase Build Team added a comment - Build couchbase-server-7.0.0-4090 contains indexing commit 48c0808 with commit message: MB-36980 force recovery on flush abort
            Hide
            Permalink
            mihir.kamdar Mihir Kamdar added a comment -

            Deepkaran Salooja can you please help with details on how to validate this issue?

            Show
            mihir.kamdar Mihir Kamdar added a comment - Deepkaran Salooja can you please help with details on how to validate this issue?
            Hide
            Permalink
            deepkaran.salooja Deepkaran Salooja added a comment -

            We were able to verify from the system test logs that this fix was able to get indexer out of the deadlock situation after flush abort.

            Show
            deepkaran.salooja Deepkaran Salooja added a comment - We were able to verify from the system test logs that this fix was able to get indexer out of the deadlock situation after flush abort.

              People

              Assignee:
              deepkaran.salooja Deepkaran Salooja
              Reporter:
              deepkaran.salooja Deepkaran Salooja
              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