[SunHELP] ps with Solaris 10 zones

[Stevens, Adrian]

Hello,

Long time reader/first time poster.

We have a SunFire v880 with 32GB RAM and 8 processors.  Recently, it was
live upgraded from Solaris 9 (with the latest 9_Recommended.zip) to plain
Solaris 10.  If we have one or more additional (non-global) Solaris 10 zones
running, the ps command slows down, but only in the global zone.

The results of `time ps` in the global zone are:
real    0m28.533s
user    0m0.011s
sys     0m0.238s
yet, `time ps -z global` from within the global zone:
real    0m0.050s
user    0m0.011s
sys     0m0.038s
This shows a remarkable difference between using the -z argument and not
using -z in the global zone.  The above timings show no significant change
if one, two, three, four or five zones are running (in addition to the
global).

`time ps` from within another-zone
real    0m0.022s
user    0m0.005s
sys     0m0.018s
`time ps -z another-zone` from within another-zone:
real    0m0.288s
user    0m0.010s
sys     0m0.026s
This shows not much difference using the -z argument when in a zone.
Obviously you can't do `ps -z global` from within a non-global zone.

Just out of interest, doing `time ps -z another-zone` from within the global
zone:
real    0m9.526s
user    0m0.012s
sys     0m0.106s

Doing a truss on the ps command shows loads of "Err#2 ENOENT" when it's
looking at the /dev tree (which kind of makes sense since not everything has
something attached).  I suspect this is what is slowing it down, as I can't
see anything else in the truss output.

My worry is that the ps command is often used in scripting and may not have
the -z option in it.  This could slow down these scripts quite a bit if
running in the global zone.

Has anybody got an ideas?  Anybody come across this before?  Is it something
I've done?

Any ideas would be appreciated.

Regards



Adrian Stevens
Senior Networked Systems Consultant
Innovation North - Faculty of Information and Technology
Leeds Metropolitan University
England


To view the terms under which this email is distributed, please go to http://disclaimer.leedsmet.ac.uk/email.htm