Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
5.3.1
-
None
-
None
Description
I am using a single broker configured to accept both tcp and ssl connections using the configuration below
<transportConnectors>
<transportConnector name="openwire" uri="tcp://0.0.0.0:61616"/>
<transportConnector name="ssl" uri="ssl://0.0.0.0:61617"/>
</transportConnectors>
The clients are configured to use the failover transport over ssl to connect to a single broker (e.g. URL: failover:ssl://Host:61616). Upon startup the client correctly connects to the broker on port 61617 via ssl. I then simulate a network outage and the transport is interrupted and waits for the connection to become available. After a short amount of time the Broker detects inactivity on the connection and then drops the connection (viewed via JMX and Jconsole). When the network connection is restored, the client reconnects to broker on port 61616 using tcp instead of ssl on port 61617 (not good if you require ssl encryption).
It appears that the ordering of the transportConnectors in the activemq.xml file affects which transport is used upon resumption of an interrupted failover transport. It turns out that upon reconnection the first defined transport will always be used regardless of what transport was specified in the original client connection url. Changing the ordering to the following fixes the problem and upon reconnection the ssl transport will be used. However, the reverse issue will occur for clients that originally connection via tcp transport where upon reconnection they will connect via ssl transport instead of tcp.
<transportConnectors>
<transportConnector name="ssl" uri="ssl://0.0.0.0:61617"/>
<transportConnector name="openwire" uri="tcp://0.0.0.0:61616"/>
</transportConnectors>