We rolled Collect out on our Tru64 5.0a and Tru64 5.1 servers last summer.
I am very pleased with the tool, but recently we have started to see
instances where Collect will use up 200% of the CPU (as shown by ps aux) on
the box for a period of time. These appear to occur exclusively on our
Peoplesoft servers. I have been able to isloate the start time as being
when it writes the data file to disk (daemon.log: Feb 12 06:06:00 server
/usr/sbin/collect[18578]: received SIGUSR1: forcing data buffer flush), and
the stop time is typically the next flush. These events seemed to happen
much more often (multiple times daily) after the PeopleTools were upgraded
from 8.14 to 8.16. To further complicate issues, two of the servers with
this issue were upgraded from 5.0a to 5.1, and one is now running PK3 and
the other PK4. We are using Collect 2.01. Has anyone else seen this kind
of problem? Any clever troubleshooting techniques?
Thanks
Bill Skulley
Received on Tue Feb 12 2002 - 20:15:49 NZDT