Uploaded image for project: 'Couchbase Server'
  1. Couchbase Server
  2. MB-47489

Backport - Eventing : Retry DcpGetSeqnos on failure

    XMLWordPrintable

Details

    • Bug
    • Status: Open
    • Major
    • Resolution: Unresolved
    • 6.6.0
    • 6.6.6
    • eventing
    • 1

    Description

      As seen in the linked CBSE, it is possible that due a transient issue while getting vb high seq numbers from KV, we skip creating DCP streams for all vbuckets owned by that eventing consumer instance.

      Adding a retry mechanism will alleviate such a scenario.

      Attachments

        Issue Links

          For Gerrit Dashboard: MB-47489
          # Subject Branch Project Status CR V

          Activity

            Build couchbase-server-7.0.2-6503 contains eventing commit 4b222de with commit message:
            MB-47489 : Retry DcpGetSeqnos on failure

            build-team Couchbase Build Team added a comment - Build couchbase-server-7.0.2-6503 contains eventing commit 4b222de with commit message: MB-47489 : Retry DcpGetSeqnos on failure

            Steps to reproduce

            1. Setup a cluster with 1 KV nodes and 1 Eventing node
            2. We need to simulate a scenario of bad connection between Data service (memcached process) and Eventing process. One way of doing so is to add firewall rule to block requests going to port 11210.
            3. Deploy an eventing function with feed boundary "From Now".

            Observation without fix:

            Eventing function deploys even after a failed attempt at opening up DCP streams and failed checkpointing.

            Observation with the fix:

            Eventing function will remain in "deploying" state until the connection to KV node is back to normal.

            abhishek.jindal Abhishek Jindal added a comment - Steps to reproduce 1. Setup a cluster with 1 KV nodes and 1 Eventing node 2. We need to simulate a scenario of bad connection between Data service (memcached process) and Eventing process. One way of doing so is to add firewall rule to block requests going to port 11210. 3. Deploy an eventing function with feed boundary "From Now". — Observation without fix: Eventing function deploys even after a failed attempt at opening up DCP streams and failed checkpointing. Observation with the fix: Eventing function will remain in "deploying" state until the connection to KV node is back to normal.

            People

              abhishek.jindal Abhishek Jindal
              abhishek.jindal Abhishek Jindal
              Votes:
              0 Vote for this issue
              Watchers:
              3 Start watching this issue

              Dates

                Created:
                Updated:

                Gerrit Reviews

                  There is 1 open Gerrit change

                  PagerDuty