DSXchange

I have a PX Job that is to upsert records to a DB2 table. The job consists of a sequential file which goes through a transformer and then the DB2 stage which performs the upsert.

The job will insert around half the records and then hang (typcially 1.5 million rows at this point). But, maybe 2 times out of 20 tries it successfully processed all 3 million rows into the table??

Does this sound like an issue with the nodepools being specified? or perhaps the database connections???

Any suggestions are appreciated...

Neal
Do you have an estimate of update to insert ratio?? If it is small try splitting into two streams (Update and insert). How wide is your table??

Try to monitor other DB2 activities and other resources.
Thanks

..It will be mostly updates after the initial load has taken place.

But, I should of stated that when encountering the issue I have truncated the table between runs to force it to just inserts and the problems still exists. THanks Neal

Hello nalombardi,

what is your commit frequency? Does the hang behaviour change if you change this? When a hang occurs, does the UNIX process use 0% CPU or does it disappear? Doe the hang always occur at the same place (either row or wall-clock time)? If the wall-clock time is the same, then you are getting a timeout of some sort...