Uploaded image for project: 'Couchbase Server'
  1. Couchbase Server
  2. MB-46668

FTS: Failover/rebalance/add-back does not cleanup indexes

    XMLWordPrintable

Details

    • Bug
    • Status: Open
    • Major
    • Resolution: Unresolved
    • Cheshire-Cat
    • backlog, Neo.next
    • fts
    • None
    • Untriaged
    • 1
    • Unknown

    Description

      Build: 7.0.0 build 5251

      • Create a 2 node cluster with kv and fts on each node
      • Load travel-sample
      • Create a default index on travel-sample with 1 partition
      • Failover and rebalance search node
      • Note that pindex files not cleaned in failed over node
      • Now drop the bucket travel-sample
      • Add-back same search-node with search service.
      • We see the indexes created before failover still exists even without source bucket

      Logs:

      https://cb-jira.s3.us-east-2.amazonaws.com/logs/failover/collectinfo-2021-06-02T022804-ns_1%40172.23.97.211.zip
      https://cb-jira.s3.us-east-2.amazonaws.com/logs/failover/collectinfo-2021-06-02T022804-ns_1%40172.23.97.212.zip

      Attachments

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

        Activity

          During failover, we won't clean up the index files from the node. This is done to make sure that the index files are reusable upon recovery of the failed-over node.

          I don't see this as a significant issue rather just a way the system behaves. 

          The user never deleted the index for the index to go away. The bucket was deleted when the FTS node was not a part of the cluster so the node won't be able to do anything with that.

          Once it is recovered/added back, it would reload the index at the bootup time.

          The user can delete the index and move on.

           

          We could also have logic during index bootup to throw away indexes whose source buckets/collections no longer exist.

          But not a very high priority at the moment.

           

           

           

          Sreekanth Sivasankaran Sreekanth Sivasankaran added a comment - During failover, we won't clean up the index files from the node. This is done to make sure that the index files are reusable upon recovery of the failed-over node. I don't see this as a significant issue rather just a way the system behaves.  The user never deleted the index for the index to go away. The bucket was deleted when the FTS node was not a part of the cluster so the node won't be able to do anything with that. Once it is recovered/added back, it would reload the index at the bootup time. The user can delete the index and move on.   We could also have logic during index bootup to throw away indexes whose source buckets/collections no longer exist. But not a very high priority at the moment.      

          The ideal thing to do is "We could also have logic during index bootup to throw away indexes whose source buckets/collections no longer exist." If it comes to "The user can delete the index and move on.", then it should be documented.

          Understood that this is not very high priority as this is not critical flow.

          girish.benakappa Girish Benakappa added a comment - The ideal thing to do is "We could also have logic during index bootup to throw away indexes whose source buckets/collections no longer exist." If it comes to "The user can delete the index and move on.", then it should be documented. Understood that this is not very high priority as this is not critical flow.

          People

            Sreekanth Sivasankaran Sreekanth Sivasankaran
            girish.benakappa Girish Benakappa
            Votes:
            0 Vote for this issue
            Watchers:
            3 Start watching this issue

            Dates

              Created:
              Updated:

              Gerrit Reviews

                There are no open Gerrit changes

                PagerDuty