RE: SQL performance problem on one job only

["Elvis Budimlic" <ebudimlic@xxxxxxxxxxxxxxxxxxxxxxxxx>]

Data skew?  Bad data?  Not enough main memory in the pool?  Someone deleting
key indexes or keyed LFs?

-----Original Message-----
Subject: SQL performance problem on one job only

Hi guys

Anyone help me with this before I scream, please?
I have one 3rd Party job that contains 30 embedded SQL commands that runs ok
for a while (runtime between 10 secs and 3 minutes in batch) and then,
inexplicably, runs for over an hour without ever finishing.

So far, I've
1) Checked DBMON for missing logicals. Created some but this seem to make it
worse.
2) Checked for damaged logicals. Replaced them all anyway
3) Recompiled program.
4) Converted program from OPM to ILE and recompiled it.
5) Created copy of program for just that (BPCS) environment.
6) Triple checked all PTFs / Cumpacks / Feng Shui
7) Checked with supplier / internet.  "The job is functioning as designed"
8) Archived data  (it's a General ledger job)
9) Amended QAQINI (or whatever it's called) settings. 
10) Removed some logicals that weren't in other environments - no effect.
11) Checked PRTSQLINF - learnt nothing from this.
12) Swore a lot

I have found references to problems with progs over 16mb in size and this
prog is over 16mb.
Currently the only workaround that works is to have it a scheduled job to
recompile it every two hours. My gut feeling is that the fact that this
workaround seems to work proves that the 16mb issue is the crux of the
problem.  It's not the only prog above 16mb but it does appear to be the
only one with the problem.

BTW This is on a 720 running V5R2.

I've run out of solutions other that rewrite the prog.

Regards

Martin Bath