I have already raised couple of post about Datastage job hanging issue and this is also about the same. I was chasing this issue with IBM from Datastage 8.1 - 8.7 and yesterday, finally i managed to talk with IBM Engineering dept and get the conclusion!
Issue: As mentioned in the subject, PROD Datastage PX jobs are hanging in random and we don't know what is the cause? Few days before a job was hanging (which was doing upsert -550 records. It inserted 350 records successfully and there after no response!!) for few days and i allowed it to trace the root cause.
After a good number of communications with IBM i managed to install necessary STACK trace, relevant ENV variable, prior to this issue!!.
Code: Select all
I have provided all the necessary logs to IBM and they came back to me, saying it is bug in the product and especially in Windows OS .
At this stage they dont know why it is causing this issue, but they suspect it could be because of tsort (internal) / sort stage !! IBM is blaming MKS Toolkit and they are working to rid out of MKS. I thing Ray was saying from 9.1 there is no MKS Tool kit. IBM too confirmed in 9.1 they have taken it out from the CLIENT and not in the SERVER!!!
Code: Select all
The workaround is to disable the parallelism for tsort alone!!
So end of the story is, a bug in the product and it took years for me to make them agree!