[AMQ-3673] Unexpected behavior in FailoverTransport when hosts are unknown

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 5.5.0
• Fix Version/s: 5.6.0
• Component/s: JMS client

Environment

MacOS X 10.5
Java(TM) SE Runtime Environment (build 1.6.0_26-b03-384-10M3425)

• Patch Info: Patch Available
• Assignee: Timothy A. Bish
• Reporter: Stefan Riesen
• Created: 20/Jan/12 16:37
• Updated: 23/Jan/12 20:47
• Resolved: 23/Jan/12 20:47

Description

There's an unexpected behavior when a failover url is specified where the first host does not exist, e.g.

failover:(tcp://nonexistinghost.mydomain.com:61616,tcp://existinghost.mydomain.com:61616)?randomize=false

The failover code will never switch to the tcp://existinghost.mydomain.com:61616 while when using this url:

failover:(tcp://nonexistinghost.mydomain.com:61617,tcp://existinghost.mydomain.com:61616)?randomize=false

The failover works flawlessly. I tracket the problem down to the method org.apache.activemq.transport.failover.FailoverTransport.contains(URI):

private boolean contains(URI newURI) {

        boolean result = false;
        try {
        for (URI uri:uris) {
            if (newURI.getPort()==uri.getPort()) {
                InetAddress newAddr = InetAddress.getByName(newURI.getHost());
                InetAddress addr = InetAddress.getByName(uri.getHost());
                if (addr.equals(newAddr)) {
                    result = true;
                    break;
                }
            }
        }
        }catch(IOException e) {
            result = true;
            LOG.error("Failed to verify URI " + newURI + " already known: " + e);
        }
        return result;
    }

That only resolves hostnames if the ports are not equal (this is why the different port behavior comes from).

In above mentioned case in the second call to this method this line fails with a UnknownHostException:

InetAddress addr = InetAddress.getByName(uri.getHost());

and the result of contains() is set to true because of the catch block, therefore the second URL isn't added to the list of urls.

I suggest to change the catch block to return false. In case of an IOException during the check it's hard to say whether the URIs are the same or not, but it's more likely that they're not the same as in this case.