Details

    • Sub-task
    • Status: Closed
    • Major
    • Resolution: Duplicate
    • None
    • None
    • None
    • None

    Description

      We need new implementation of snapshot feature that is more robust and performant. Ideally, it will work with multiple tables as well. The possible areas of improvements:

      1. Must be flushless. Coordinated memstore flushes across a cluster are bad.
      2. Verification phase must be distributed, done in parallel and not on Master.

      In theory, the only info we need to record snapshot of a table: list of WAL files, list of HFiles and max sequence id of an edit which has been flushed per Region.

      Attachments

        Issue Links

          Activity

            People

              Unassigned Unassigned
              vrodionov Vladimir Rodionov
              Votes:
              0 Vote for this issue
              Watchers:
              7 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved: