[
https://issues.apache.org/jira/browse/AMQ-3811?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Buchi Reddy B updated AMQ-3811:
-------------------------------
Description:
Hi,
I am running ActiveMQ brokers with Shared file system Master/Slave
configuration and I am seeing the following issue.
1. Start Master and slave brokers.
2. Start a durable STOMP consumer (Perl/Python) and a producer (Perl/Python)
with persistence.
3. Kill the master broker while the producer is still sending messages.
4. Our STOMP library takes care of failing over to the new broker automatically
and the consumer replays all its subscriptions. So, producer continues to send
to the slave broker and consumer also receives all the messages from slave
broker.
5. Stop the producer and consumer.
6. Restart the consumer with same client-id and subscription name.
After this, I see that the consumer received some messages which were already
delivered.
This is easily reproducible with the above mentioned steps. I can clearly see
in the broker logs that the slave broker has received acks from the consumer
but for some reason it doesn't purge these messages and thinks that the
consumer didn't ack them. The messages which were sent to the master broker
(before it was killed) were actually getting removed and are not delivered
multiple times.
Any help here is greatly appreciated.
Thanks a lot!!
was:
Hi,
I am running ActiveMQ brokers with Shared file system Master/Slave
configuration and I am seeing the following issue.
1. Start a durable STOMP consumer (Perl/Python) and a producer (Perl/Python)
with persistence.
2. Kill the master while the produce is still sending messages.
3. Our STOMP library takes care of failing over to the new broker automatically
and also the consumer will reply all the subscriptions. So, producer starts
sending to the new broker and consumer also receives all the messages.
4. Stop the producer and consumer.
5. Restart the consumer with same client-id and subscription name.
After this, I see that the consumer received some messages which were already
delivered. This is easily reproducible with the above mentioned steps. I can
clearly see in the broker logs that the slave broker has received acks from the
consumer but for some reason it doesn't purge these messages and thinks that
the consumer didn't ack them. The messages which were sent to the master broker
(before it was killed) were actually getting removed and are not delivered
multiple times.
Any help here is greatly appreciated.
Thanks a lot!!
> Slave broker in the Master/Slave configuration doesn't process acks of
> running STOMP consumer and redelivers all the messages if the consumer
> restarts
> ------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: AMQ-3811
> URL: https://issues.apache.org/jira/browse/AMQ-3811
> Project: ActiveMQ
> Issue Type: Bug
> Components: Broker, stomp
> Affects Versions: 5.5.1
> Environment: Linux operating system and Perl/Python STOMP clients.
> Reporter: Buchi Reddy B
>
> Hi,
> I am running ActiveMQ brokers with Shared file system Master/Slave
> configuration and I am seeing the following issue.
> 1. Start Master and slave brokers.
> 2. Start a durable STOMP consumer (Perl/Python) and a producer (Perl/Python)
> with persistence.
> 3. Kill the master broker while the producer is still sending messages.
> 4. Our STOMP library takes care of failing over to the new broker
> automatically and the consumer replays all its subscriptions. So, producer
> continues to send to the slave broker and consumer also receives all the
> messages from slave broker.
> 5. Stop the producer and consumer.
> 6. Restart the consumer with same client-id and subscription name.
> After this, I see that the consumer received some messages which were already
> delivered.
> This is easily reproducible with the above mentioned steps. I can clearly see
> in the broker logs that the slave broker has received acks from the consumer
> but for some reason it doesn't purge these messages and thinks that the
> consumer didn't ack them. The messages which were sent to the master broker
> (before it was killed) were actually getting removed and are not delivered
> multiple times.
> Any help here is greatly appreciated.
> Thanks a lot!!
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
