Red Hat Bugzilla – Bug 489004
Newly joined node may end up with duplicate enqueue events on replication queue
Last modified: 2015-11-15 19:07:00 EST
State for replication queue will be transfered to the newly joining node therefore event generation should be suppressed when enqueing messages on other queues as part of the catchup.
Fixed by r751719.
Test procedure (e.g.):
(1) start primary cluster node (node_1),
(2) declare queue with --generate-queue-events 2,
(3) send messages to this node,
(4) consume half of those messages (again through connecting to this node)
(5) now add second node to that cluster (node_2),
(6) start a dr broker and establish a bridge from node_2 to that broker,
(7) check that replication to dr broker works and that only expected messages are on the replicated queue
There is a test checked into qpid svn (clustered_replication_test) that may be useful and is run during make check.
Tested on r750054 and it didn't work as I expected.
Testend on r752581 and it works. clustered_replication_test passed.