Uploaded image for project: 'HBase'
  1. HBase
  2. HBASE-26193

Do not store meta region location as permanent state on zookeeper

    XMLWordPrintableJSON

    Details

    • Type: Improvement
    • Status: Resolved
    • Priority: Major
    • Resolution: Fixed
    • Affects Version/s: None
    • Fix Version/s: 2.5.0, 3.0.0-alpha-2
    • Component/s: meta, Zookeeper
    • Labels:
      None
    • Hadoop Flags:
      Reviewed
    • Release Note:
      Hide
      Introduce a new 'info' family in master local region for storing the location of meta regions.
      We will still mirror the location of meta regions to ZooKeeper, for backwards compatibility. But now you can also clean the meta location znodes(usually prefixed with 'meta-region-server') on ZooKeeper without mess up the cluster state. You can get a clean restart of the cluster, and after restarting, we will mirror the location of meta regions to ZooKeeper again.
      Show
      Introduce a new 'info' family in master local region for storing the location of meta regions. We will still mirror the location of meta regions to ZooKeeper, for backwards compatibility. But now you can also clean the meta location znodes(usually prefixed with 'meta-region-server') on ZooKeeper without mess up the cluster state. You can get a clean restart of the cluster, and after restarting, we will mirror the location of meta regions to ZooKeeper again.

      Description

      As it breaks one of our design rules

      https://hbase.apache.org/book.html#design.invariants.zk.data

      We used to think hbase should be recovered automatically when all the data on zk (except the replication data) are cleared, but obviously, if you clear the meta region location, the cluster will be in trouble, and need to use operation tools to recover the cluster.

      So here, along with the ConnectionRegistry improvements, we should also consider move the meta region location off zookeeper.

        Attachments

          Issue Links

            Activity

              People

              • Assignee:
                zhangduo Duo Zhang
                Reporter:
                zhangduo Duo Zhang
              • Votes:
                0 Vote for this issue
                Watchers:
                10 Start watching this issue

                Dates

                • Created:
                  Updated:
                  Resolved: