I decided this is something best opened up to this august group. I think
I may have a
hardware/OS or configuration problem.
I have a script which labels all of the elements in all of the vobs
selected
by a normal config spec. The script was originally written to spawn a
separate process for each vob. When we ran it, it would complete in
about
30-45 minutes. Some time in the last year, the time increased to an hour
or more, and, when the script was running, performance for anyone
accessing
clearcase became impossibly slow. The machine running the script is an
SGI
Origin 2000, with 12 CPU's. The short term solution was to change the
script to
run each vob serially. The performance problem went away, but the
labeling took
3 hours. Yesterday, I decided to test an older version of the script. It
forced the
user to run it on a specific machine, a small sun box (Ultra 2, I
think). The older
version of the script ran all of the vobs in parallel. I ran three
passes over the course
of about 2 hours. The three passes took 16minutes, 12 minutes, and
20minutes
respectively, and there were no noticeable performance problems.
Does anyone out there have any kind of an explanation for this much of a
change?
Could it be that the 12cpu machine is just overwhelming the view/vob
server combination
with data (ie its too fast). Or could this be some kind of significant
configuration
problem with my big IRIX box?
Thanks for any suggestions any of you might have.
Mike
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
This archive was generated by hypermail 2b29 : Sun May 06 2001 - 00:26:28 EDT