Red Hat Bugzilla – Bug 985869
Do not make durable subscriptions auto-delete by default
Last modified: 2015-04-14 09:46:47 EDT
Description of problem:
Subscription queues defined using amqp1.0 are *not* auto-delete by default.
Subscription queues defined using amqp0-10 are auto-delete by default.
This puts amqp0-10 and amqp1.0 out of sync.
After discussion with gsim we agreed that defining durable subscription queues to be auto-delete by default without a delayed time for the auto-delete, would not make much sense.
ie. on broker restart, the broker terminates all the connections (and sessions), so in fact the durable queue is deleted on broker restart.
note: By specifying that a subscription is durable, the client is taking on responsibility for deleting it to some extent.
The right solution might be to have a configurable time-delayed auto-delete on subscription queues by default. This will protect the broker from clients that go off and leave stale subscriptions after a crash, but it still retain the general ability to survive a crash and re-establish subscription without lost messages. I'm going to fill a RFE bug for this improvement.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. $cppapi/drain -f amq.direct &
2. qpid-config queues
the just created subscription queue is auto-del
Subscription queue is auto-delete by default (amqp0-10)
Subscription queue is not auto-delete by default (amqp0-10)
> ... I'm going to fill a RFE bug for this improvement.
please see bug 985870
Fixed upstream: https://svn.apache.org/r1504621
This improvement has been implemented.
Verified on rhel6.6 (x86_64 and i386) and rhel7 (x86_64).
The timed auto-delete (default to 120s) is now added by default to the subscription queue when a durable subscription is requested. This works for both amqp0-10 and amqp1.0, but only if C++ client is used (Using other clients the auto-delete-timeout must be explicitly set for the link). please see bug 985870 for details.
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.