that needs an enhancement and may be tricky to implement b/c network connectors cannot operate properly over failover reconnects. masterslave does not do any retries - it just picks from the url list, and should respect randomize=false on the first connect and reconnect (when initiated by the networkConnector discovery agent) but because it does not do reconnects via the failover transport (maxReconnect=0) it does not track restarts and priority.
On 7 February 2014 15:56, Geurt Schimmel <gschim...@schubergphilis.com> wrote: > Hi, > > randomize=false and priorityBackup are not working with the masterslave > protocol - or is this by design ? > > A client-broker has a masterslave setup towards hostA and hostB. > > Would like to failover to hostB if ActiveMQ on hostA is not available, but if > hostA becomes available again, would like to switch back to hostA. After > starting the client-broker, I see that it's often connected to hostB - > apparently, the order I specified is not honoured, so randomize=false is not > working. When shutting down the destination-broker it is currently connected > to, there is a failover indeed, but after that broker comes back, there is no > priorityBackup, so no switch back to hostA. > > <networkConnector name="sbpamq102_sbpahdb" userName="system" > password="${activemq.password}" > uri="masterslave:(ssl://ipaddr_hostA:61616?wireFormat.maxInactivityDuration=0& > > connectionTimeout=3000,ssl://ipaddr_hostB:61616?wireFormat.maxInactivityDuration=0& > > connectionTimeout=3000)?timeout=5000&randomize=false&priorityBackup=true&maxReconnectAttempts=-1" > duplex="false" networkTTL="10" prefetchSize="3000" staticBridge="true"> > <staticallyIncludedDestinations> > ... > </staticallyIncludedDestinations> > </networkConnector> > > And the listeners on hostA and hostB: > > <transportConnector name="openwire" > uri="ssl://0.0.0.0:61616?jms.prefetchPolicy.queuePrefetch=4800" > updateClusterClients="true" updateClusterClientsOnRemove="true"/> > > I checked https://issues.apache.org/jira/browse/AMQ-4501, reason why > rebalanceClusterClients in not in the transportconnector-config. > > Thanks, > Geurt -- http://redhat.com http://blog.garytully.com