activemq-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Chris Hoffmann (JIRA)" <j...@apache.org>
Subject [jira] [Created] (AMQCPP-363) failover: consumer does not always restart properly after reconnect
Date Wed, 20 Apr 2011 17:17:06 GMT
failover: consumer does not always restart properly after reconnect
-------------------------------------------------------------------

                 Key: AMQCPP-363
                 URL: https://issues.apache.org/jira/browse/AMQCPP-363
             Project: ActiveMQ C++ Client
          Issue Type: Bug
          Components: CMS Impl
    Affects Versions: 3.3.0
         Environment: GNU Linux
            Reporter: Chris Hoffmann
            Assignee: Timothy Bish
            Priority: Critical
             Fix For: 3.3.1


Hi,

While testing AMQCPP-355, we came across a problem of when resuming from a broker restart.
Resuming seems to be erratic when the broker shutdown happens during message processing in
an onMessage callback.

a) if a message is ack'ed while the broker is still down, the same message will be redelivered
after connection resumes. The command tracing does not show the sending of the 1st acknowledgment.

b) if a message is ack'ed after the broker is brought back (connection resumed OK), the command
tracing shows that the ack is send, but on the broker side we see an warning message:

 WARN | Ignoring ack received before dispatch; result of failover with an outstanding ack.
Acked messages will be replayed if present on this broker. 
Ignored ack: MessageAck {commandId = 22, responseRequired = false, ackType = 2, consumerId
= ID:hostname-44923-1303316424844-0:0:0:0,
firstMessageId = ID:hostname-58084-1303317597544-0:0:0:0:0,
lastMessageId = ID:hostname-58084-1303317597544-0:0:0:0:0,
destination = queue://test, transactionId = null, messageCount = 1, poisonCause = null}

We can't see any redelivery message in the command trace and subsequent messages are not delivered
either. The consumer is nevertheless shown on broker admin console. 

When the connection is resumed, we see that the prefetch size is set to 0 in the consumerInfo
message.

c) resuming works fine, if the message is not in process when the shutdown happens.

Cases a+b work fine, when we tried to reproduce with a java client from the ActiveMQ examples
dir.


Regards,
Chris

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message