Why suddenly longer startup times?

Hesubo · Post by **Hesubo** » Wed Feb 20, 2008 1:11 pm

For the past several days the startup time on many of our jobs has increased from approx. 30 seconds to over 2 minutes. We are, of course, trying to find out why. DBA, Unix Admins, network admins and ETL programmers all state that no changes have been made to the system.

On suggestion from IBM is to add the following to our jobs to gain more information: APT_STARTUP_STATUS, APT_PM_PLAYER TIMING, APT_DUMP_SCORE, OSH_DUMP.

How can I estimate how much additional time those things will add to our run times?

I do not believe the OSH_DUMP will give us information that we here can understand. We don't speak Osh. Will the information provided by the other settings give us information that will make sense to developers who are not experts in performance tuning or should I plan on getting a performance tuning expert to anlayze the additional info.?

Also, if you ahve any ideas or experience regarding why startup times would increas as they have here, I would appreciate your input.

Thanks,
Bob

Environment is AIX MPP, 4 Lpars with 8 CPUs each/DS 7.5.2/DB2/

kcbland · Post by **kcbland** » Wed Feb 20, 2008 1:50 pm

1. Clear out the &PH& directory on the server node. This is a classic issue.
2. Job logs could be very full, impeding stage startup/wrapup operations. Consider a mass purge if this is the case.
3. Monitor the server node resources using topas. If the server node is bottlenecked, EVERYTHING will startup/wrapup slow.

ray.wurlod · Post by **ray.wurlod** » Wed Feb 20, 2008 4:41 pm

Monitor, with APT_STARTUP_STATUS, both in the good times and the bad. Try to discover any correlation with the machines' workloads.