| Author |
Message
|
| malammik |
 Posted: Wed Mar 21, 2007 8:17 am Post subject: Publishing to Broker on Remote Queue Manager |
 |
|
Partisan
Joined: 27 Jan 2005
Posts: 397
Location: Philadelphia, PA
|
JMS Client is connected to queue manager A, however topic destination has a BROKER QUEUE MANAGER parameter pointing to queue manager B which is happily talking to A and vice versa.
1. In the even of the broker failure on QM B, would the JMS client get an exception?
2. Is it possible to publish under a transaction to a broker on a remote queue manger, if so where would the message be sitting waiting for a commit? On subscriber queues?
Broker is WMBI 6.0 + latest CSDs. MQ is 6.0 with latest CSDs
I will be doing a lab exercise to try these out shortly myself but any prelimiary feedback is as always greatly appreciated.
_________________
Mikhail Malamud
http://www.netflexity.com
http://groups.google.com/group/qflex |
|
| Back to top |
|
 |
| malammik |
| Posted: Wed Mar 21, 2007 9:41 am Post subject: Here is the answer to my first question |
|
|
Partisan
Joined: 27 Jan 2005
Posts: 397
Location: Philadelphia, PA
|
|
| Back to top |
|
|
| jefflowrey |
| Posted: Wed Mar 21, 2007 9:49 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
The transaction should be held against the transmit queue.
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| malammik |
| Posted: Wed Mar 21, 2007 9:56 am Post subject: |
|
|
Partisan
Joined: 27 Jan 2005
Posts: 397
Location: Philadelphia, PA
|
That's what I am thinking. BUT but.....! Consumer has a notion of receive isolation which implies there might be uncommitted messages sitting on the subscriber queue hence the suspicion above that the uncommitted messages might be sitting on the subscriber queues. In other words, if the publication is under a transaction, broker will drive the message as far as it can (to subscriber subscribe queue) uncommitted.
_________________
Mikhail Malamud
http://www.netflexity.com
http://groups.google.com/group/qflex |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Wed Mar 21, 2007 10:03 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
Broker can't process the publication messages until they're committed.
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| malammik |
| Posted: Wed Mar 21, 2007 1:41 pm Post subject: |
|
|
Partisan
Joined: 27 Jan 2005
Posts: 397
Location: Philadelphia, PA
|
So what you're saying is that when I commit the publication, the actual pub message has not even been processed as such by the broker. I have to agree with yo based on my knowledge but would like to think that the JMS and the broker are smarter than that.
_________________
Mikhail Malamud
http://www.netflexity.com
http://groups.google.com/group/qflex |
|
| Back to top |
|
|
| malammik |
| Posted: Wed Mar 21, 2007 1:45 pm Post subject: |
|
|
Partisan
Joined: 27 Jan 2005
Posts: 397
Location: Philadelphia, PA
|
Keep publishing on my own thread. Perhaps this will help someone else down the road. Consider this scenario.
PUBACKINT is default at 25 which means JMS client expects an acknowledgement from the broker every 25th message published. Now if we publish 26 message under a transaction, our transaction should then fail because broker did not even receive the 1st message yet since it is uncommitted on the xmitq. I will test that out tonight and let you know.
_________________
Mikhail Malamud
http://www.netflexity.com
http://groups.google.com/group/qflex |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Wed Mar 21, 2007 1:52 pm Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
No, I mean this is basic MQ transactionality.
The publication message is not available to the broker until it's committed. So until the publisher commits, the broker can't perform the publication, as the publication hasn't really happened.
I do not believe there is any way to control the syncpoint between broker and subscriber queue. Depending on which pub/sub broker you use, it's likely to either always be in syncpoint or syncpoint-if-persistent.
Then it is up to the subscriber to determine if they need a syncpoint for themselves, when getting the publication message. But the broker has already considered that the message is published by this point.
I think in your PUBACKINT case, the publisher will still get the ack.
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
|
|
