Clemens did a great session yesterday afternoon on SOA troubleshooting. He gave a nice overview of the threading and transaction architecture of BPEL and ESB.
He addressed the famous "lost instance" issue with BPEL, which can happen when errors occur inside a BPEL process before it's had a chance to persist itself to the database. Of most interest to me is the fact that in 10.1.3.4, the dehydration of the audit trail data will occur asynchronously in a second JTA transaction. The only issue we have with going to 10.1.3.4 for some of our clients is it is missing some important bug fixes from the 10.1.3.3.1 branch of the code. Clemens mentioned that MLR 1 (for 10.1.3.4) is coming soon to resolve this issue.
0 comments:
Post a Comment