Description
In https://issues.apache.org/jira/browse/ARTEMIS-2462 we offered an option that automatically remove snf queues in scaledown. However the remote bindings are left out and remains in the broker memory.
Those are no longer used and if a different node comes up those remaining bindings will prevent the new bindings to be added.
For a common example in a 2 broker cluster,
if they both deploy a jms.queue.DLQ queue, each will have a remote binding
for the queue. One of them scaled down and remove the sf queue on the other.
Then another broker node (with different node id) comes up and form a cluster with the existing broker. If the new broker also has a jms.queue.DLQ then it will cause
the other broker to create a remote binding. However the other broker already has a remotebinding and for that reason the new remotebinding will not be added.
You will see some warning like this:
2019-09-12 01:30:51,427 WARN [org.apache.activemq.artemis.core.server] AMQ222139: MessageFlowRecordImpl [nodeID=a44b0e0a-d4fc-11e9-9e65-0a580a8201d0, connector=TransportConfiguration(name=artemis, factory=org-apache-activemq-artemis-core-remoting-impl-netty-NettyConnectorFactory) ?port=61616&host=ex-aao-ss-2, queueName=$.artemis.internal.sf.my-cluster.a44b0e0a-d4fc-11e9-9e65-0a580a8201d0, queue=QueueImpl[name=$.artemis.internal.sf.my-cluster.a44b0e0a-d4fc-11e9-9e65-0a580a8201d0, postOffice=PostOfficeImpl [server=ActiveMQServerImpl::serverUUID=f04e96b2-d4fc-11e9-8b50-0a580a8201d2], temp=false]@522d336c, isClosed=false, reset=true]::Remote queue binding DLQf04e96b2-d4fc-11e9-8b50-0a580a8201d2 has already been bound in the post office. Most likely cause for this is you have a loop in your cluster due to cluster max-hops being too large or you have multiple cluster connections to the same nodes using overlapping addresses
Attachments
Issue Links
- is part of
-
ARTEMIS-2462 Allow store and forward queue to be deleted after scaledown
- Closed