I have a Hub-Sub setup for DB2 on MVS. I have a Data Migrator flow which selects records from DB2, joins them to a local SQL*Server table, and then writes the records back to DB2 with changes.

The Problem -- DB2 crashes the run after performing a small number of updates. Not a dependable number, sometimes the first record crashes it, sometimes the 400th. But it never runs to completion short of me requesting an exceptionally small set of records. It seems to be based on the record itself, some records jamming the system consistently while others sail through.

When I set the Commit number to a small number, say 10, when the run is in a position to run the next 50 records successfully, the job hangs at 10 records, updating the first ten but never going back to update more, and never returning control of DMC to me.

My INSERTSIZE is set to 1, my FASTLOAD is currently set off, but doesn't seem to have any effect.

I opened a ticket but it became very complicated to get IBI the information they needed to proceed and I have about a dozen other fires burning.

Has anybody seen anything like this before?

This message has been edited. Last edited by: <Emily McAllister>,

I think you should put this question up in the iWay forum. Not all WebFOCUS folks are DM users.

It's in both. There is exactly one user in the I-Way forum at the moment, and that's me. I hoped to get some extra eyes on it by posting here.

Are you writting to SUFFIX=DB2 or SUFFIX=EDA?

I've tried both with no change in behavior.

Writing to a SUFFIX=EDA wouldn't work anyway, but it would explain what you are seeing. You should only be writing to a defined data source. So, check out the DB2 adapter on HUB side. I assume that you are using the DB2 CLI or JDBC. In either case, you shouldn't these problems unless the account doesn't have write access to the DB2 instance.

I assume that DB2 is located with the SUB server. You should verify that you can write to DB2 from the SUB server. Worst case scenario is you break the flow into two parts. One to extract the DB2/MS SQL query to a flat file (if the output is not expected to be massive). The second to write to the final DB2 destination. If you have to, execute the second step via -REMOTE BEGIN/END so that it runs directly on the SUB server. Be very careful with the defining meta-data in that case ... you can put yourself into an endless loop very easily if you are not careful.

SUFFIX=EDA worked as well as SUFFIX=DB2, but both seem to hang up on particular records.

I think I'm going to go after this via StarSQL on Win2K8 instead, avoiding the MVS iWay server entirely. Might have better luck.