Uploaded image for project: 'Couchbase Server'
  1. Couchbase Server
  2. MB-7522

Optimize very common index state transitions (rebalance)


    • Type: Improvement
    • Status: Resolved
    • Priority: Critical
    • Resolution: Fixed
    • Affects Version/s: 2.0.1
    • Fix Version/s: 2.0.1
    • Component/s: view-engine
    • Security Level: Public
    • Labels:
    • Flagged:
      Release Note


      Optimize 2 types of state transitions:

      1) If a state transition consists only of moving partitions
      between the passive and active states (or vice-versa), that
      is, no new partitions are added nor existing partitions are
      marked for cleanup, don't restart the updater (at the expense
      of more complex logic to correct snapshots received from a
      currently running updater).

      2) If a state transition only adds new partitions to the
      passive state (as ns_server's rebalance frequently calls),
      don't restart the updater, and send the updater the new
      list of passive partitions, so that it has a chance to
      iterate over those partitions' content and index them in
      the same run.

      As long as the rebalancer in ns_server doesn't keep changing
      its logic regarding state transitions all the time, this
      provides a significant reduction of the indexing time. However
      for at least some scenarios, there's still significant portion
      of rebalance time not spent on indexing nor index compaction.

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



          • Assignee:
            FilipeManana Filipe Manana (Inactive)
            FilipeManana Filipe Manana (Inactive)
          • Votes:
            0 Vote for this issue
            0 Start watching this issue


            • Created:

              Gerrit Reviews

              There are no open Gerrit changes