The question was:
> Last Saturday a supposedly minor change was made to an application (custom
> built plus an Oracle database) running on a DS10. Concurrently with that
> change, the system CPU time shot up alarmingly. Clearly they changed
> _something_ in the application that caused this, but, trying to be helpful
> and also cover the OS end of things, is there any way to troubleshoot
what's
> causing this? Aside from killing processes and seeing when it stops?
As luck would have it, the app folks figured out what happened. They added
support for some new cranes, which required more semaphores than before, and
sem_ume was too low. Apparently their code's response to an error is to
loop. sheesh; what's that remark about programmers, civilization and
woodpeckers?
ANYway, Bryan Williams, Thomas Sjolshagen, and Tom Smith all suggested using
DCPI, and
Tim Cutts suggested using trace or truss, to track down this sort of thing,
and I'll look into those tools for next time.
Thanks all.
- Bluejay Adametz
Do not meddle in the affairs of dragons, because you
are crunchy and taste good with ketchup.
Received on Fri Feb 21 2003 - 12:44:42 NZDT