On Sep 16, 2009, at 09.03 , F J Rutcho wrote:
We're trying to take advantage of the CPU & Memory monitoring
capabilities in Spectrum under Device Thresholds. However, we're
encountering problems because our customers want to monitor at
multiple levels (e.g., Major & Critical). They like the fact that
we can set a sustain rate on a per server basis, but they want
multiple levels of monitoring (presently Spectrum only provides one).
I've also noticed that Memory monitoring can be accomplished using
multiple levels if we take advantage of the "File Systems (2790)"
under System Resources. Drawback here though is there's no place to
set a sustain rate.
We're trying to break away from using MOM for CPU & Memory
monitoring. I know SpectroWatches are an option however again
there's the issue with multiple levels and no sustain rate.
There are examples in the documentation of creating a watch that will
only trigger if a variable has been over value for N minutes. Take a
look at p118 and 119 in doc 0919 - you need an on demand watch that is
checking if an attribute is within alarm state and outputting either 1
or 0, and a polled watch that takes care of the duration problem.
It's really clever. There are also some examples of stacking together
2 or 3 watches to get an escalating degree of alarms as the problem
worsens, so you have a minor alarm at 50-75%, then a major at 75-95%,
and a critical over 95%.
I think you could combine the two sets of examples to do exactly what
you want fairly easily.
Mathias Wegner
ISC Networking
University of Pennsylvania
---
To unsubscribe from spectrum, send email to [email protected] with the body:
unsubscribe spectrum [email protected]
