JMS on Weblocig8.1 cluster

[janhes]

I am running DSTX 6.7.1 on Win2K with the JMS adapter against a Weblogic8.1 cluster with 8 nodes.
I have found that the only way I can consume from all 8 nodes is to have 8 instances of the map running at the same time with each map configured to access a specific node.
If the maps do not all run at the same time not all the messages are consumed. It appears that even though a map is configured to access a specific node, it doesn't do that.
If I then run all the maps together, the missing messages are proceessed.
Does anyone have any experience of weblogic clusters and if so how do you read all the messages from the different nodes?
Jan

 

[BocaBurger]

How do you tell the nodes apart? I never had an issue with this, but for most adapters and clusters, DS TX only "sees" one object, such as an MQ queue. Same with clustered DBs, all DS TX connects to is the local client.

BocaBurger
<===========================||////////////////|0
The pen is mightier than the sword, but the sword hurts more!

 

[janhes]

I specify the urls of the 8 nodes in the command line. It's the only way I've been able to connect.
So I specify
-ictxfurl t3://wgos308:7001,wgos308:7003,wgos309:7001,wgos309:7003,wdox276:7001,wdox276:7003,wdox275:7001,wdox275:7003 -ICTXF weblogic.jndi.WLInitialContextFactory -qn RM_ResponseQueue -cfn BatchServicesConnectionFactory -hdr -prp -lsn 10 -qty 100 -T+

 

[BocaBurger]

OK, so each map has a different URL? Give each trace a unique name (best bet use the url: 308-7001.trc for example, run the test, send the traces and the explanation to Ascential Support. This will need to go to engineering for debugging.

So if you only have one map watching one URL, it won't get any messages?





BocaBurger
<===========================||////////////////|0
The pen is mightier than the sword, but the sword hurts more!

 

[janhes]

If I only have one map watching 1 url, the results vary depending on how the Q is populated. If I run as responses are being posted I get more messages than it I run the map after all the messages have been posted. I think this is because the load balancing causes more messages to be placed on node 1 when I run as messages are posted - the number of messages on the node will stay small as I delete them.

 

[BocaBurger]

Is this running under the Event Server? Are the maps multi-threaded, or is there a resource conflict? What is your maxthreads setting, watchmaxthreads? How many CPUs on the DSTX server?

What do you see in the Event Server monitor? Management console for pending maps?

As I said before, I feel this needs to be sent to engineering. There should be no reason that one map can get messages independantly and when there are multiple maps, each URL should not have a relationship with the other, unless the messages are part of a group, and some are not being committed on the queue until the group is complete.

If that is the case, it is not a DS TX issue.



BocaBurger
<===========================||////////////////|0
The pen is mightier than the sword, but the sword hurts more!

 

[janhes]

Apparently the problem is because the adapter consumes the messages as transacted. As a result only one node is accessed. I think by running 8 consumers together the WebLogic load balancing forces the maps to access different nodes.

 

[janhes]

Sent this to engineering but they didn't know enough about weblogic clustering to give an answer.