Uploaded image for project: 'Couchbase Server'
  1. Couchbase Server
  2. MB-35560

Swap rebalance failed for KV due to premature auto-failover

    XMLWordPrintable

Details

    • Untriaged
    • Unknown

    Attachments

      Issue Links

        For Gerrit Dashboard: MB-35560
        # Subject Branch Project Status CR V

        Activity

          Build couchbase-server-6.5.0-4652 contains ns_server commit 7b3c543 with commit message:
          MB-35560 do not failover node when the node doesn't see the bucket

          build-team Couchbase Build Team added a comment - Build couchbase-server-6.5.0-4652 contains ns_server commit 7b3c543 with commit message: MB-35560 do not failover node when the node doesn't see the bucket

          Build couchbase-server-6.5.0-4652 contains ns_server commit 9ada04c with commit message:
          MB-35560 store bucket UUIDs in buckets_with_data to prevent

          build-team Couchbase Build Team added a comment - Build couchbase-server-6.5.0-4652 contains ns_server commit 9ada04c with commit message: MB-35560 store bucket UUIDs in buckets_with_data to prevent

          Build couchbase-server-7.0.0-1019 contains ns_server commit 7b3c543 with commit message:
          MB-35560 do not failover node when the node doesn't see the bucket

          build-team Couchbase Build Team added a comment - Build couchbase-server-7.0.0-1019 contains ns_server commit 7b3c543 with commit message: MB-35560 do not failover node when the node doesn't see the bucket

          Build couchbase-server-7.0.0-1019 contains ns_server commit 9ada04c with commit message:
          MB-35560 store bucket UUIDs in buckets_with_data to prevent

          build-team Couchbase Build Team added a comment - Build couchbase-server-7.0.0-1019 contains ns_server commit 9ada04c with commit message: MB-35560 store bucket UUIDs in buckets_with_data to prevent
          dfinlay Dave Finlay added a comment -

          Moving this ticket to Cheshire Cat. With the work that's already happened it should be the case that we don't see spurious auto-failovers for in freshly added nodes participating in rebalance. We'll consider the issue of memcached buckets in Cheshire Cat. The reasons for deferring are:

          1. I believe the risk as it relates to memcached buckets are acceptable. We haven't seen these spurious auto-failovers commonly at all and memcached buckets are less commonly seen than Couchbase buckets.
          2. The work to address memcached buckets is relatively sophisticated and for reasons of risk I'd like to do this work not in Mad Hatter
          3. We are considering taking a fresh look at how we consider when nodes are candidates for failover.
          dfinlay Dave Finlay added a comment - Moving this ticket to Cheshire Cat. With the work that's already happened it should be the case that we don't see spurious auto-failovers for in freshly added nodes participating in rebalance. We'll consider the issue of memcached buckets in Cheshire Cat. The reasons for deferring are: I believe the risk as it relates to memcached buckets are acceptable. We haven't seen these spurious auto-failovers commonly at all and memcached buckets are less commonly seen than Couchbase buckets. The work to address memcached buckets is relatively sophisticated and for reasons of risk I'd like to do this work not in Mad Hatter We are considering taking a fresh look at how we consider when nodes are candidates for failover.

          People

            artem Artem Stemkovski
            mahesh.mandhare Mahesh Mandhare (Inactive)
            Votes:
            0 Vote for this issue
            Watchers:
            5 Start watching this issue

            Dates

              Created:
              Updated:

              PagerDuty