Re: [tkined] Questions about my code design. . . opinions please?

pdebono@eur.ko.com
Wed, 27 Jan 1999 14:55:21 +0100

Hi Robert,

I had a similar project, to poll x no. of (Bay) routers and various
switches (C100 & Accelar). I tried several approaches, but settled on a
"wrapper" script to determine the device type then spawn a device specific
process. Spawned process determines interfaces & configurations etc. and
then takes relevant action to look for potential error conditions specific
to each device & it's interface type / speed etc.

This probably wasn't the best way to do it, but it allowed for additional
device types to be easily added and could handle device configuration
changes in our very dynamic environment. This also allowed for some very
remote devices who's responses can be slow to not affect overall polling
and program execution speed. The "wrapper" script limits the no. of spawned
processes active at any one time. I also prioritise device scheduling such
that key devices are polled more often than lower priority ones. This is
necessary as it would take too long to poll them all sequentially. I have
several hundred components being monitored this way now without problems.

I have to carefully limit the number of devices being checked at any one
time (I limit to 5). This is more to do with not wanting to flood the WAN
with SNMP than any limitation of TNM or the host system that I have come
across. System resources have not been a problem so far, but that would
probably be host dependant anyway. The same AIX server has other unrelated
TNM scripts running simultaneously.

This doesn't really answer your questions, and I doubt I am qualified to do
so anyway. However, I have a need to increase both the number of devices
being polled and the amount of checking done on some of the devices. The
approach I have used is going to take too long to execute if I do this, so
I am also keenly interested in the advice of the list.

Peter Debono.
Technical Consultant,
The Coca-Cola Company.

--
!! This message is brought to you via the `tkined & scotty' mailing list.
!! Please do not reply to this message to unsubscribe. To subscribe or
!! unsubscribe, send a mail message to <tkined-request@ibr.cs.tu-bs.de>.
!! See http://wwwsnmp.cs.utwente.nl/~schoenw/scotty/ for more information.