activemq-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mark Gellings (JIRA)" <jira+amq...@apache.org>
Subject [jira] Commented: (AMQNET-294) durable subscription message loss when master broker fails to slave
Date Wed, 15 Dec 2010 17:50:01 GMT

    [ https://issues.apache.org/jira/browse/AMQNET-294?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12971756#action_12971756
] 

Mark Gellings commented on AMQNET-294:
--------------------------------------

As far as your confusion...the basic issue is with a durable subscription, upon failing a
master broker to the slave, message consumption stops.  Only until I restart the message consumer
do the remaining messages come through.  When the issue was originally created I did not restart
the consumer after the broker fail over (as shouldn't need to), so I assumed message loss
occurred.  Whether the inactivity monitor is enabled or not, the messages do not come through
when a broker fails over until I restart the consumer.

My latest testing is with ActiveMQ 5.4.2 and ActiveMQ NMS trunk.  I am using JDBC Master/Slave,
the brokers run on Windows Server 2008 and consumer on Windows 7.

The way I can replicate this issue is when physically failing over the broker or by using
your test case.  

The only differences when I run your test case is the broker is not running locally and I
am using JDBC master/slave.

We can create a new issue if you would like.  For now I will run a broker locally and run
your test to see if it passes.  

One additional detail is I do see the following message in the broker log file after I fail
over the broker if that means anything to you.

2010-12-15 11:26:25,147 | INFO  | Transport failed: java.io.IOException: Wire format negotiation
timeout: peer did not send his wire format. | org.apache.activemq.broker.TransportConnection.Transport
| Async Exception Handler

> durable subscription message loss when master broker fails to slave
> -------------------------------------------------------------------
>
>                 Key: AMQNET-294
>                 URL: https://issues.apache.org/jira/browse/AMQNET-294
>             Project: ActiveMQ .Net
>          Issue Type: Bug
>          Components: NMS
>    Affects Versions: 1.4.1
>         Environment: Windows 7 (client), Windows Server 2008 64-bit (server brokers run
on), Sql Server 2008 (database)
>            Reporter: Mark Gellings
>            Assignee: Jim Gomes
>             Fix For: 1.5.0
>
>         Attachments: Apache.NMS.Test (2).zip, Apache.NMS.Test.zip, DurableConsumerTest.cs,
DurableSubscriberFailoverTest.java, TimsTestRevisedSlightly.zip
>
>
> We are seeing message loss on a durable subscription when using NMS ActiveMQ v1.4.1 and
ActiveMQ v5.4.1.
> Please run the included NUnit test and watch the console output.  When it says "Failover
the broker now!" do as it says.  About 75% of the time less than half of the expected 250
messages come through.
> Using version 1.1 of NMS the majority of the time the test passes.  I have seen it fail
only a few times with this earlier version, and when it does there are only a couple messages
that don't come through.
> In the zip file will be the unit test, and a config directory containing the master and
slave activemq configurations.  We are using JDBC master/slave.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message