Uploaded image for project: 'Cassandra'
  1. Cassandra
  2. CASSANDRA-6455

Improve concurrency of repair process

    XMLWordPrintableJSON

Details

    Description

      Currently, most of the repair tasks (taking snapshots, send/receiving merkle tree, compute MT difference, etc) are done on single threaded AntiEntropyStage.

      This causes a problem like CASSANDRA-6415 and likely to cause unnecessary wait.

      Also, repair is done one CF at the time. I think we can parallelize this(concurrency is configurable by a user based on # of CF and load of the nodes) for faster processing.

      Attachments

        1. 6455.txt
          145 kB
          Yuki Morishita
        2. 6455-3.0.txt
          146 kB
          Yuki Morishita
        3. 6455-3.0-v2.txt
          175 kB
          Yuki Morishita
        4. 6455-tinyfix.patch
          1 kB
          Marcus Eriksson

        Issue Links

          Activity

            People

              yukim Yuki Morishita
              yukim Yuki Morishita
              Yuki Morishita
              Marcus Eriksson
              Votes:
              2 Vote for this issue
              Watchers:
              16 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved: