SUMMARY: E10K/Solaris 7 106541-21 performance problem

From: Swider, Bucky <jswider_at_concordefs.com>
Date: Wed Mar 26 2003 - 14:21:32 EST
Thanks to:

Tom Crummey
Jeff Horwirtz
Darren Dunham
Sebastian Daubigne
Justin Stringfellow
Kevin Buterbaugh

Who pointed out the rather blatant presumption I had screwed up.  I need to
do a <-f> in my truss command to trace child processes.  I had figured that
since I didn't see the fork, that it was waiting on something
internally....but since I started the truss midstream, the fork had happened
before the waitid.  I'll try again tonight to catch this if the problems
reoccurs.  Thanks for all the fast responses!

Bucky

> 
> Hello, we have a mission-critical app whose window is compromised due to
> poor performance.  I TRUSSed one of the jobs last night and found that the
> process was sleeping.  I suspect some sort of lock condition....I'm
looking
> for some jumping-off points to start my investigation.  The code in
question
> is a ksh script.  Here's the smoking gun:
> 
> truss -p 27994
> waitd(P_ALL, 0, 0xFFBEE018, WEXITED|WTRAPPED|WSTOPPED) (sleeping...)
> .
> <at least thirty minutes passes with all resources on system idle>.....
> .
> .
> .
> waitd(P_ALL, 0, 0xFFBEE018, WEXITED|WTRAPPED|WSTOPPED) = 0
> 
> <job then flies to completion within seconds>.
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
Received on Wed Mar 26 14:28:02 2003

This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:43:07 EST