Uploaded image for project: 'AMQ Broker'
  1. AMQ Broker
  2. ENTMQBR-7532

Failovers in 2-way Federated HA Broker Pairs Result in Phantom or Missing Consumers

XMLWordPrintable

    • False
    • None
    • False
    • Hide
      • Configure a shared-store ha broker pair using the attached configuration. Configure 20 queues for each broker
      • Configure a second shared-store ha broker pair separated from the first pair over a network with the same queue configuration as the first pair
      • Configure an upstream queue federation on each broker in the 2 pairs
      • Start up the master and backup brokers in each pair.
      • Start 1 (ha) consumer for each queue on each end of the federated cluster, for a total of 40 consumers (20 on each end)
      • Stop and restart the primary broker at each end of the cluster in turn, checking the consumer count on the live brokers at each end after each failover.
      • Eventually you see a state where one of the live brokers is reporting extra consumers or missing consumers.
      • This persists until the next failover event.
      Show
      Configure a shared-store ha broker pair using the attached configuration. Configure 20 queues for each broker Configure a second shared-store ha broker pair separated from the first pair over a network with the same queue configuration as the first pair Configure an upstream queue federation on each broker in the 2 pairs Start up the master and backup brokers in each pair. Start 1 (ha) consumer for each queue on each end of the federated cluster, for a total of 40 consumers (20 on each end) Stop and restart the primary broker at each end of the cluster in turn, checking the consumer count on the live brokers at each end after each failover. Eventually you see a state where one of the live brokers is reporting extra consumers or missing consumers. This persists until the next failover event.

      In a scenario where 2 shared-store ha broker pairs are connected via federation (for example 2 datacenters with one pair in each datacenter), failing over the brokers by stopping and restarting the master in each pair can result in orphan consumers. For example, in a scenario with 20 local consumers attached to each ha pair, the cluster correctly reports 40 consumers at each end of the federated broker cluster. After stopping and restarting the primary broker at each end, we might see 60 consumers reported at one or the other end of the federated cluster. The count persists until the remote broker (from the one showing phantom consumers) is failed over once again.

        1. artemis-server-2.21.0.redhat-00041.jar
          2.22 MB
        2. ENTMQBR-7532.zip
          75.30 MB
        3. ENTMQBR-7532-tweaked.zip
          75.68 MB
        4. reproducer-1.zip
          17 kB

              rhn-support-jbertram Justin Bertram
              rhn-support-dhawkins Duane Hawkins
              Samuel Gajdos Samuel Gajdos
              Votes:
              0 Vote for this issue
              Watchers:
              3 Start watching this issue

                Created:
                Updated:
                Resolved: