Hide Forgot
Description of problem: Messaging User Guide (staging version) states, that limit policy flow_to_disk "can apply to both persistent and non-persistent queues" (see http://documentation-stage.bne.redhat.com/docs/en-US/Red_Hat_Enterprise_MRG/2/html/Messaging_User_Guide/chap-Messaging_User_Guide-Queues.html "Controlling queue size"). Nevertheless, if the command 'qpid-config add queue --limit-policy=flow_to_disk' is called without the '--durable' option, then limit policy of the resulting queue is set to 'reject': [root@lzhaldyb-rhel62x ~]# qpid-config add queue test --max-queue-count=100 --limit-policy=flow-to-disk --flow-stop-count=0 --flow-resume-count=0 [root@lzhaldyb-rhel62x ~]# echo $? 0 qpidd.log: trace RECV [[::1]:5672-[::1]:48687]: Frame[BEbe; channel=1; {QueueDeclareBody: queue=test; arguments={qpid.flow_resume_count:F8:int64(0),qpid.flow_stop_count:F8:int64(0),qpid.max_count:F8:int64(100),qpid.policy_type:V2:12:str16(flow_to_disk)}; }] info Queue "test": Policy created: type=reject; maxCount=100; maxSize=104857600 Although qmf shows the limit policy for such a queue as 'flow_to_disk': [root@lzhaldyb-rhel62x ~]# qpid-config queues Queue Name Attributes ============================================================ test --max-queue-count=100 --limit-policy=flow-to-disk it uses the 'reject' policy instead: [root@lzhaldyb-rhel62x messaging]# ./spout -c 101 test 2012-02-01 17:45:18 warning Exception received from broker: resource-limit-exceeded: resource-limit-exceeded: resource-limit-exceeded: Policy exceeded on test, policy: size: max=104857600, current=0; count: max=100, current=100; type=reject (qpid/broker/QueuePolicy.cpp:87) [caused by 105 \x00:\x00] 2012-02-01 17:45:18 warning Exception received from broker: resource-limit-exceeded: resource-limit-exceeded: resource-limit-exceeded: Policy exceeded on test, policy: size: max=104857600, current=0; count: max=100, current=100; type=reject (qpid/broker/QueuePolicy.cpp:87) [caused by 1 \x00:\x00] Failed to connect (reconnect disabled) If the 'flow_on_disk' limit policy is in fact not applicable to non-durable queues, than this should be changed into documentation bug with request to alter chapter 3 of MIG accordingly. Version-Release number of selected component (if applicable): python-qpid-0.14-1.el6.noarch python-qpid-qmf-0.14-3.el6.x86_64 qpid-cpp-client-0.14-1.el6.x86_64 qpid-cpp-client-devel-0.14-1.el6.x86_64 qpid-cpp-client-devel-docs-0.12-6.el6.noarch qpid-cpp-server-0.14-1.el6.x86_64 qpid-cpp-server-devel-0.14-1.el6.x86_64 qpid-cpp-server-store-0.14-1.el6.x86_64 qpid-cpp-server-xml-0.14-1.el6.x86_64 qpid-java-client-0.14-1.el6.noarch qpid-java-common-0.14-1.el6.noarch qpid-java-example-0.14-1.el6.noarch qpid-qmf-0.14-3.el6.x86_64 qpid-tools-0.14-1.el6.noarch How reproducible: always Steps to Reproduce: 1. Try to create a non-durable queue with limit policy 'flow_to_disk' (f.e., by running "qpid-config add queue test --max-queue-count=100 --limit-policy=flow-to-disk --flow-stop-count=0 --flow-resume-count=0"). 2. Try to send to the newly created queue an amount of messages which exceeds max-queue-count set by the previous command. Actual results: The limit policy 'reject' was used instead, the exception 'resource-limit-exceeded' received from the broker. Expected results: The queue was created with the given limit policy 'flow_to_disk', no exceptions from the broker, the messages exceeding max-queue-count limit are saved to disk. Additional info:
The flow-to-disk mode only works on durable queues. The documentation quoted above is incorrect. Only durable queues have an active journal associated with them, and this is needed to write and recover flowed messages. When no store is present for a queue, there is no option available to a queue but to reject the message. Note that a distinction must be made between the persistence of queues and of the messages on that queue. If a queue is non-persistent, then none of the messages will be persisted, no matter the persistence status of each message. However, if a queue is persistent, then the persistence status of the message determines whether a message will be persisted. For flow-to-disk messages, when the limit is reached, both persistent and non-persistent messages will be stored. The above documentation link is stale. It is possible that the documentation is meant to say "can apply to both persistent and non-persistent _messages_". I recommend opening a documentation bug. It is also possible that the queue logic that handles the policy should warn if flow-to-disk policies are used on non-persistent queues at the time the queue is declared. I recommend changing this bug to CLOSED - NOTABUG.
I'm changing this into a documentation bug. The MCIG chapter "4.9.1. Controlling Queue Size" contains the description of flow-to-disk policy without stating that it is applicable to durable queues only. The information provided by Kim in comment 2 should be added to the documentation.
Added notes "effective only on durable queues" and "the messages stored to disk are not persistent unless the message is marked persistent." http://deathstar1.usersys.redhat.com:3000/builds/7069-Messaging_Installation_and_Configuration_Guide/#Controlling_Queue_Size