Uploaded image for project: 'Kafka'
  1. Kafka
  2. KAFKA-155

Support graceful Decommissioning of Broker

    Details

    • Type: Improvement
    • Status: Resolved
    • Priority: Major
    • Resolution: Fixed
    • Affects Version/s: 0.7
    • Fix Version/s: 0.8.0
    • Component/s: None
    • Labels:
      None

      Description

      There should be a graceful way of decommissioning the broker so that there is absolutely 0 data loss. Decommissioning is not necessarily related to replication (Kafka-50).

      There should be a way to get the broker out of the cluster only from the produce side. Consumers should be able to continue keep pulling data. When the administrator is sure that all data has been consumed by consumers, broker node can be removed permanently.

      Same would be useful for rolling upgrades without any message loss.

        Activity

        Hide
        brugidou Maxime Brugidou added a comment -

        I think this is done in 0.8 branch with the ShutdownBroker admin command (which goes through JMX command KafkController.shutdownBroker())

        Show
        brugidou Maxime Brugidou added a comment - I think this is done in 0.8 branch with the ShutdownBroker admin command (which goes through JMX command KafkController.shutdownBroker())
        Hide
        nehanarkhede Neha Narkhede added a comment -

        ShutdownBroker admin command does not completely achieve this. It merely moves the leader from that broker and then shuts the broker down. But now, some of the partitions could be under replicated. What really solves the problem is preferred replica election admin command. This allows you to add new brokers to an existing cluster, move some partition off of the brokers to be decommissioned and then shutdown the broker by killing it.

        Show
        nehanarkhede Neha Narkhede added a comment - ShutdownBroker admin command does not completely achieve this. It merely moves the leader from that broker and then shuts the broker down. But now, some of the partitions could be under replicated. What really solves the problem is preferred replica election admin command. This allows you to add new brokers to an existing cluster, move some partition off of the brokers to be decommissioned and then shutdown the broker by killing it.
        Hide
        junrao Jun Rao added a comment -

        You mean the partition reassignment tool, not preferred replication election tool, right?

        Show
        junrao Jun Rao added a comment - You mean the partition reassignment tool, not preferred replication election tool, right?
        Hide
        nehanarkhede Neha Narkhede added a comment -

        Oh, right, i meant the partition reassignment tool

        Show
        nehanarkhede Neha Narkhede added a comment - Oh, right, i meant the partition reassignment tool

          People

          • Assignee:
            Unassigned
            Reporter:
            sharadag Sharad Agarwal
          • Votes:
            1 Vote for this issue
            Watchers:
            8 Start watching this issue

            Dates

            • Created:
              Updated:
              Resolved:

              Development