[j-nsp] Graphing VCP Backplane (possible jnxUtil SNMP bug?)

[Charlie Allom]

On Thu, Nov 03, 2011 at 12:39:44PM -0400, Phil Shafer <phil at juniper.net> wrote:
> Charlie Allom writes:
> >I've come across another performance issue - running through the jnxUtil
> >MIB - it drives the load up significantly:
> 
> The two key questions are is the script completing before you launch
> the next invocation (use jcs:dampen() to limit this) and is the CPU
> spent in "cscript" (the process running the op script) or in
> snmpd/mib2d (doing the snmp operations)?
> 
> Some notes:
> 
> Looks like you're still using jcs:invoke() inside fetch-object().

Cheers Phil,

Doh! So I am! Thanks for all the tips .. I've tidied the script and am
now running jcs:execute. I'll fiddle with the rest later on, but for
now, it seems like my load has dropped totally. I ran it overnight and
there was some small spike which might be the timing issue you are
mentioning (I'll look into jcs:dampen()).

In answer to your question, the script was sitting mostly in mgd(?) with
some mib and snmp nonsense. 

5292 root        1 132    0 35664K 26012K RUN      0:01 31.00% mgd
5194 root        1   4    0  5380K  3020K sbwait   0:00  0.00% cscript

More here: http://capslock.playlouder.com/show.system.processes.extensive.txt

One thing I've noticed is that I have to copy the script to both op and
event locations. I'm sure I just haven't read enough yet to understand
why..

set event-options generate-event check-vc-throughput time-interval 300
set event-options policy check-vc-throughput events check-vc-throughput
set event-options policy check-vc-throughput then event-script check-vc
set event-options event-script file check-vc.slax
set system scripts op file check-vc.slax

I've put the script here with some graphing config for anyone else that
is interested in their VC throughput: https://github.com/yeled/check-vc

Regards,
  C.
-- 
 +442077294797
 http://mediasp.com/