So it seems you have four brokers in your network of brokers: A1, B1, A2,
B2. They are all live (that is, none are part of a master/slave pair where
the slave is idle) with duplex connections like this
A1 <----> B1, B1 <----> B2, B2 <-----> A2, A2 <----> A1.
How many messages are "getting lost"? Where are the producers and where are
the consumers connecting?
Why is TTL set to 10?
Have you enabled debug logging to see what the brokers are doing? Are
messages getting stuck on some brokers? Or bouncing between brokers? Are
all messages being sent persistent? Only using Queues? What does the rest
of the networkConnector configuration look like (statically included
destinations, etc)?
On Wed, Mar 27, 2013 at 4:14 AM, Geurt Schimmel <
gschim...@schubergphilis.com> wrote:
> Hi,
>
> We have 2 network-of-brokers with a failover-configuration, but we're
> having performance-problems and loosing messages. Clients connect to HostA1
> and HostA2. All brokers run 5.8 stable, only persistent messages using
> KahaDB:
>
> <persistenceAdapter>
> <kahaDB directory="${activemq.base}/data/kahadb"
> enableJournalDiskSyncs="false" indexWriteBatchSize="10000"
> indexCacheSize="1000"/>
> </persistenceAdapter>
>
> HostA1:
> uri="static:(ssl://HostA2:61616)" duplex="true"
> uri="static:(failover:(ssl://HostB1:61616,HostB2:61616)?randomize=false&priorityBackup=true)"
> conduitSubscriptions="false" decreaseNetworkConsumerPriority="true"
> duplex="true" networkTTL="10" prefetchSize="4000" staticBridge="true"
> suppressDuplicateQueueSubscriptions="true"
>
> HostA2:
> uri="static:(failover:(ssl://HostB2:61616,ssl://HostB1:61616)?randomize=false&priorityBackup=true)"
> conduitSubscriptions="false" decreaseNetworkConsumerPriority="true"
> duplex="true" networkTTL="10" prefetchSize="4000" staticBridge="true"
> suppressDuplicateQueueSubscriptions="true"
>
> HostB1:
> uri="static:(ssl://HostB2:61616)" duplex="true"
>
> HostB1 and HostB2:
> uri="ssl://
> 0.0.0.0:61616?jms.prefetchPolicy.queuePrefetch=4000&jms.optimizeAcknowledge=true"
> updateClusterClients="true" rebalanceClusterClients="true"
> updateClusterClientsOnRemove="true"/>
>
> Note the host-order in the URI is mirrored.
>
> When the brokers on HostA2 and HostB2 are down, performance is amazing (>
> 1000 msg/sec), and all messages are received by HostB1.
>
> Please advise - I'm lost.
>
> Geurt
>
>
--
*Christian Posta*
http://www.christianposta.com/blog
twitter: @christianposta
