We have ActiveMQ clustering between our two servers configured as a network
of brokers. Our config files look like
<broker xmlns="http://activemq.apache.org/schema/core";
brokerName="receiver" persistent="false" useJmx="false"
dataDirectory="${activemq.base}/data"
destroyApplicationContextOnStop="true">
<networkConnectors>
<networkConnector
uri="static:(ssl://le2.comt:61616<http://le1.liveexpert.net:61616/>
)"/>
</networkConnectors>
and
<networkConnectors>
<networkConnector
uri="static:(ssl://le1.comt:61616<http://le1.liveexpert.net:61616/>
)"/>
</networkConnectors>
We're starting to see some errors in our JBoss logs that look like
14:53:37,812 INFO
[org.apache.activemq.transport.failover.FailoverTransport] (ActiveMQ
Task-55) Successfully connected to ssl://le2.com:61616?keepAlive=true
14:54:30,213 INFO
[org.apache.activemq.transport.failover.FailoverTransport] (ActiveMQ
Task-57) Successfully connected to ssl://le2.com:61616?keepAlive=true
I took a look at the source for
org.apache.activemq.transport.failover.FailoverTransport - this message
gets written to log when a successfully reconnection event has been made
from the client - in this case, Java/JBoss. Since JBoss is configured with
jms_url as a static list of brokers, i.e.,
failover:(
ssl://le1.com:61616?keepAlive=true,ssl://le2.com:61616?keepAlive=true)
JBoss attempts to connect to one of the brokers upon startup, the
connection logic randomizes which broker to connect to by default. What I
think is happening when these messages get written is the client has lost
connection to the broker it has connected to initially, due to hitting the
inactivityTimeoutThreshold of 30 seconds somehow - and is attempting to
reconnect to one of the nodes in our network of brokers. Is this a correct
assumption? Is it correct that the Java code running in our JBoss
container is writing these types of 'reconnect' messages to the logs? We
see these messages about once a day, is that too frequent for that period
of time? The reason I ask is that some clients are reporting errors, and
we were able to correlate that to when these 'reconnecton' events appear in
our logs.
The reason the client needs to reconnect to the failover cluster is unclear
to me. There are no indications of an ActiveMQ server restart at this
time, just that connection between client and broker needs to be
reestablished. Can anyone recommend anything we can do to prevent these
types of reconnection events from happening as much as possible, given our
setup? Or anything we can reconfigure that may be causing frequent
reconnection events? Thanks.
