MQSeries.net :: View topic - Best Practices for Triggered Application Design

dotjas · Posted: Sun Nov 17, 2002 11:32 pm Post subject:

We are currently designing a Java/JMS application that will be triggered by MQ and need to know the best practices with regards to handling messages that cause exceptions.

Basic Processing Cycle:
Once the application is triggered, it will take each message, parse the XML, and update the database. Repeating until all messages have been processed. It will wait and try again for certain scenarios (as identified below) and throw messages that cause exceptions into a "Problem Queue", then keep processing.

What we have done:
All exception scenarios have been identified and we have divided them into two groups;
1. Scenarios that the application should wait and try to process the message again (e.g. database is down).
2. Scenarios that we will send the message to a "Problem Queue" and perform recovery on these messages at a later time (e.g. XML did not parse successfully, causing an exception when updating the database, etc.).

Background:
a.) Messages must be processed in FIFO order with exception of those that went to the "Problem Queue".
b.) We are using the "First" option for triggering (perhaps with a Wait Interval).
c.) Message volume is estimated to be between 1,000 - 3,000 per day.

The Question:
Is this the best design practice(s) to use for a triggered application?

I've searched through the site, but cannot find anything that is close. Any Input?

_________________
Jim Schuchardt
Principal Consultant
MQ Certified Specialist