Mercator Integration with JMS

[mlapse]

Hi,

We use mercator maps to write to a JMS queue. When we ran the system in production, we found that we were getting errors in the launcher file like (Target Not Available)

THis error was not reported for all messages, only some. I believe it could be because of the idle connections becoming invalid and mercator trying to use them.

I need to know whether there is some way to simulate this. Is there some timeout limit for Weblogic JMS connections?

Also there are parameters in the mercator.ini file. However since I cant simulate this condition (on the test machine, it seems to work no matter when i try to put the files and no matter how many files I have)

Please help!

 

[janhes]

I found a downside to adding
option2=-Xms200m
option3=-Xmx300m
to the .ini file.
The event server ran out of memory after 5 days.

 

[mlapse]

janhes,

What is your RAM size? We are using 2 GB RAM and my settings are 256 MB and 512 for the max and min heap size.

Did you get an error in your log stating "cannot allocate a heap size of <x> MB"? Maybe there are other processes which are running on your machine which caused the memory to run out. Generally when the JVM starts paging to disk the performance slows down and may result in a crash.

So what were the symptoms of your error? How did you handle it?

I am testing the fix right now and it seems to be working. Maybe Ill put it in production in a couple of days and let you know what happens.

 

[janhes]

We seem to have a memory leek in the Event Server. The memory usage exceeded the VM size and maps failed saying there wasn't enough memory to run them. The rate of memory leak is related to the number of JMS messages processed.

 

[mlapse]

Janhes,

This issue has been fixed and it has been almost two weeks since we have had the JMS stuff running on productio without any errors.

I got a fix from IBM support. They provided us with updated *.jar files. Also you need to make sure that all your dll's are on the correct version. Do a mercver and send it to IBM support. They will provide you with the latest dll's if you require them.

Also the JMS system is now using 256MB as the min and 512 MB as the max heap size.

Let me know if you need further info.

 

[janhes]

Were the Jar files m4jms.jar and m4 java.jar with classes dated 14 & 18/10/05? If so I've recently got them and am waiting to put them into production.
Did you have to set the heap size to 256MB as the min and 512 MB as the max to get it to work ok or was this just a recommendation?

 

[mlapse]

That is correct. Those were the files received from support. Mine were dated 19/10/05 because once you confirm the fix, they will package it into a release.

I set the heap sizes at various values and it seemed to work properly with these settings. Thats the reason we kept them on in production. Of course it depends on the size and volume of messages you are handling as well as the capacity of your server. If it s too high, JAVA will start paging to the disk and will cause problems for you.

Make sure all other dll's are of the same version in test as well as production before you move to prod. If there is any issue with the working of mercator, make sure you raise it to support.

Hope this helps.

 

[janhes]

Well I put the new jars in yesterday and it made no difference. In the java trace file I get
*ERR: java.lang.Exception: ****Error in cmdline or adapter properties
but from the log file there's nothing wrong with the comand line or adapter properties.
Back to the drawingboard

 

[mlapse]

This is something best handled by IBM support. I havent faced this issue.

What is the modified date on your new jar files?

 

[janhes]

Well I've finally got over this problem. The new JAR files did solve the problem. However, when we first implemented them the old jar files had been renamed m4jms.date.jar and m4java.date.jar. The trace files for JMS showed that we were still using the old JAR files. It turns out that the event server loads jar files by loading *.jar and the old JAR fiels were taking preference to the new ones. Once the renamed files were removed everything was ok.