[
https://issues.apache.org/activemq/browse/AMQ-1820?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=43688#action_43688
]
ying commented on AMQ-1820:
---------------------------
The topology we want to have is as follows:
machine A:
masterBroker1
slaveBroker2
machine B:
masterBroker2
slaveBroker1
or more machines to have pure master/slave pair
on client machine, we want to indicate only the masters:
failover://(tcp://masterBroker1:port, tcp://masterBroker2:port)
so the slaves are only used for replication purpose.
if any one broker regardless of master or slave dies, its corresponding master
or slave will finish what it is doing but not accepting new messages or
connections and all the messages will failover to the available masters.
Is this a feasible in the current release?
After all, the goal for us is to have network of brokers to have HA and slaves
only for replication so messages are not lost. Thanks
> Pure Master/Slave-kill the slave-master will not work
> -----------------------------------------------------
>
> Key: AMQ-1820
> URL: https://issues.apache.org/activemq/browse/AMQ-1820
> Project: ActiveMQ
> Issue Type: Bug
> Environment: Tested on both solaris 9 sparc and windows XP with
> release ActiveMQ 5.1 release
> Tested ActiveMQ 5.1 release on xp only.
> on solaris, master and slave are two machines. on window xp, master and slave
> are on one machine with different ports.
> Reporter: ying
> Attachments: ActiveMQ5.1PureMasterSlaveTest.zip,
> masterSlaveTestCase.zip
>
>
> Step to reproduce:
> Test 1
> 1, Setup a Pure Master/Slave (master is using the original activemq.xml,
> slave is added masterConnectorURI="tcp://masterhost:62001"
> shutdownOnMasterFailure="false"
> 2, Kill the slave
> 3, Have a client to send a message to a queue.Master will not be able to
> function, see MasterTest1.log (WARN TransportConnection - Failed
> to add Connection
> javax.jms.JMSException: The destination topic://ActiveMQ.Advisory.Connection
> and admin console queues.jsp throws exceptions too)
> Test 2
> 1. Same as Test 1
> 2. Same as Test 1
> 3. restart the slave. Master will throw exception( on solaris master will
> throw exception, slave has no exception, on xp, both will throw exception and
> it is an infinite loop so I have to kill both) See MasterTest2.log for the
> exception
> 4. Have a client to send a message to a queue.Master will not be able to
> function, same exception as slave restart is thrown.
> Both MasterTest1.log MasterTest2.log files are the tests on solaris.
> Rest is a Win XP test for Test2, which has an infinite loop
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
