Facing performance, weblogic <BEA-000337><BEA-001112> strucktheard issue while generating the XML us

Dear Members/Readers,

As we have build an Custom Utility around the Vignette API to extract the GUID's from VCM and to build/create/generate the XML's for Each content GUID's, which holds all the information like metadata,version,classifications..etc pertaining to that GUID. As we are having roughly 7039 content id in our QA env, while running our utility for entire/complete records (7039) we are able to run it for approx for 1000 -2000 odd records, with out any interruption and there after it seems ,our custom appplication/utility gets into halted/hangup stage. And nothing is been processed further.

After 30 - 35 mins of halt/idle time, the VCM server seems to be up/active and our utility picks up processing the records from where it has left. And the total time it has taken to run the utility for the entire 7039 records is approx 3 hours.

As it seems to be some form of perfromance issue, as well as the exception/error at the web logic server side also.when browsed in to the VgnVCMServer_output.log, we were able to get the below mentioned error.

--------------------------------

<Dec 10, 2013 10:08:33 PM EST> <Error> <WebLogicServer> <BEA-000337> <ExecuteThread: '23' for queue: 'weblogic.kernel.Default' has been busy for "1,061" seconds working on the request "connection144.session145", which is more than the configured time (StuckThreadMaxTime) of "600" seconds.>
<Dec 10, 2013 10:18:33 PM EST> <Error> <WebLogicServer> <BEA-000337> <ExecuteThread: '23' for queue: 'weblogic.kernel.Default' has been busy for "1,661" seconds working on the request "connection144.session145", which is more than the configured time (StuckThreadMaxTime) of "600" seconds.>
<Dec 10, 2013 10:27:43 PM EST> <Error> <JDBC> <BEA-001112> <Test "SELECT 1 FROM SYSIBM.SYSDUMMY1" set up for pool "VgnPool" failed with exception: "java.sql.SQLException: [Vignette Corp.][DB2 JDBC Driver]Connection timed out".>
<Dec 11, 2013 3:18:33 AM EST> <Error> <WebLogicServer> <BEA-000337> <ExecuteThread: '10' for queue: 'weblogic.kernel.Default' has been busy for "1,030" seconds working on the request "com.vignette.as.server.ejb.SiteOpsBean_o3ciu8_EOImpl", which is more than the configured time (StuckThreadMaxTime) of "600" seconds.>
<Dec 11, 2013 3:28:33 AM EST> <Error> <WebLogicServer> <BEA-000337> <ExecuteThread: '10' for queue: 'weblogic.kernel.Default' has been busy for "1,630" seconds working on the request "com.vignette.as.server.ejb.SiteOpsBean_o3ciu8_EOImpl", which is more than the configured time (StuckThreadMaxTime) of "600" seconds.>
<Dec 11, 2013 3:37:10 AM EST> <Error> <JDBC> <BEA-001112> <Test "SELECT 1 FROM SYSIBM.SYSDUMMY1" set up for pool "VgnPoolXA" failed with exception: "javax.transaction.xa.XAException: [Vignette Corp.][DB2 JDBC Driver]XAStart failed.".>

------------------------------------
Also attached the VCM server log for complete refference and understanding.

Would request for any quick help and assistance on the above issue.

Appreciate any form of help in the above aspect.

Warm Regards

Ganesh.

Find more posts tagged with

Comments

Evan_Bench

seems you have two choices: (1) tune your weblogic and db to accomodate these very long transaction times or (2) modify your code so that it is more 'aware' of thread-based transactions.

It seems that the custom app is trying to get all content instances in one transaction (which is alot!). can you modifiy the code to process these in chunks?

If not, then you'll want to increase distributed lock timeout in both DB and Weblogic (set db side to 60 seconds greater than that set in weblogic - this is called JTA in weblogic and distrubuted lock timeout in Oracle).

GANESHKUMAR_JEYAPRAKASH

Hi Evan,

Thanks a ton, for an quick response..we really appreciate it and it is helping us a lot.

Just to ellaborate more on the Custom app, we are getting all our Content GUID's which are imported from DB2 using another program in the same utility.So we fecthing all the 7039 GUID's from the custom table, and pass each guid to the managed object using the iterator's to get the desired name's from the managed object of VCM ( Like ManagedObjectVCMRef (id),content type..etc)

Once we have all the data from the managedobject for all the GUID's we keep them in a list and pass on to another method (In this method also we are iterating the count ), were we are processing to form/create/generate the GUID XML's for each and every content id. in the middle of this process only, the system is getting halted.

In the generatecontentxml () method only, we are trying to get all the metadata info (like classification,channels,version's,secutiry...etc) pertaining to that Content id. while processing and building the xml's, here is where we are facing the system hang up issue.

So,keeping all the facts intact, are there any other ways/approaches to reslove/overcome this issue. As the two choices you have mentioned, we dont have much control over them. Here is explaination from our side on them:

1) we can't modify, any web logic/db2 server config, as they are out of reach for us.

2) We are using an simple JDBC back end application ( not an web application )build on jdk 1.4 as this is the version the server are as well as the Vignette Supporting jars too are build on 1.4 it self.

Cheers,

Ganesh.

GANESHKUMAR_JEYAPRAKASH

Hi Evan,

Thanks a lot Evan for your valuable inputs, Could you please suggest for the mentioned scneraio what could be the ideal settings for the below option.

(1) tune your weblogic and db to accomodate these very long transaction times.

Thanks & Regards

Ganesh

Eric Borisow

Ganesh,

I don't know the application specifically, but one approach may be to simply start by taking the ids that are returned from the first method and store them somewhere (such as the file system or db). Then, modify your second method to create a threadpool and have each thread process maybe 50 content ids at a time and write the results of those somewhere. Finally, assemble the results from all of the threads into one large xml file on completion of all of the threads.

Just a suggestion.

Regards,

Eric