-
1. Re: I got failure in queue-message-redistribution
It works fine before at r6939, and it fails after r6940...
I'm taking a look -
2. Re: I got failure in queue-message-redistribution
"clebert.suconic@jboss.com" wrote:
It works fine before at r6939, and it fails after r6940...
I'm taking a look
Actually.. it aways fails. I made some confusion with ant clean/ run on my last test.
It fails this way since when the test was written. I just tried r6549 which is April/24th, and it failed the same way. -
-
4. Re: I got failure in queue-message-redistribution
I wouldn't worry too much about this for now, it's been a known issue for a while.
We'll fix it before beta, but I don't consider this #1 priority right now. -
5. Re: I got failure in queue-message-redistribution
"timfox" wrote:
It doesn't always fail for me
It *aways* fail when I run it as:
ant clean run -
6. Re: I got failure in queue-message-redistribution
I've reduced a bit the scope of the failure: it always fails if *server1/data* is deleted before the test is run.
Delete server0/data and the example still works
Starting server1 before server0 and the example still works (in this case, the deletion of server1/data will still make the example fails) -
7. Re: I got failure in queue-message-redistribution
in order to understand the problem, here is what happening when the example pass (i.e. server1/data
contains bindings and logs from a previous run):
- a consumer is created on server0
=> server1 is notified the consumer is created
- on server1, the queue has no local consumers but 5 message references
=> a redistributor is created
However, there shouldn't be 5 messages references in the queue! They come from the
previous run where a NPE occurs and there remains 5 messages in the queue.
To sum up, the example will always fail the 1st time and this failure will make the example
succeeds unless the server1/data is cleaned up.
Now, the example will systematically fail.
What I'm expecting is that a redistributor must be added to server1 when it is notified
that *its own consumer is closed*, there should be 5 messages references in the queue
(since they were not acked by the consumer) but there is none. I'm investigating why it is so.
To prevent that, I'll make the example use non persistent messages so that they won't be
any messages from a previous run.