Individual, not global, thresholds for CPU/Memory/RT

Its rather annoying that the thresholds for all the default gauges are global, global is fine but there needs to be a override for individual nodes such as in the case of SQL servers or Exchange servers that run high memory usage on purpose.

Currently the only way around this is to create a poller and get the value with SNMP. After that you have the old gauge which is sitting there doing nothing for you, and if you want to clear an alarm threshold for that node you have to bump up the thresholds on your entire environment.

Find more posts tagged with

threshold

prodfr

Status: None

Comments

Sohail.Bhamani

Many folks create custom properties to store the per device values for thresholds. These can then be referenced in the alert engine instead of using a static value or the words warning or critical. This thwack post shed great light on this:

http://thwack.solarwinds.com/thread/41978

Sohail Bhamani

Loop1 Systems

http://www.loop1systems.com

Bahlkris

Right... but then you still have a gauge that is red on that particular node, and its going to stay red till you change your threshold globally. *problem* Its a great piece of work by adatole but its alert centric. It puts you in a place where you have to do a lot of custom work while at the same time basically blowing away your global thresholds. If you want metircs on all your nodes at say "level A" with the exception of say 10 sql servers at "level B" and 5 exchange servers at "level C" then you pretty much have to setup a custom poller for every device in your environment. AND you are still throwing away the default thresholds, so why even have them in the first place? So yes, there is a way around it but it seems like it should be something that is built into the product. Hence, feature request.

FormerMember

What if, on the same switch, I expect 60% utilization on one interface but only 5% on another....

Is there a way to handle this?

It seems natural to expect a Monitoring Tool to easily allow you to set thresholds on a per object basis, simply and eloquently.

Sohail.Bhamani

One way to accomplish this would be to label the varying interfaces with a custom property. Then you can reference this custom property in the alert engine so you can specify the varying thresholds you want. It is not perfect, but it will work with today's feature set.

Sohail Bhamani

Loop1 Systems

http://www.loop1systems.com

phesse80

There are two other open feature requests for this by Phillip Hesse and Bahlkris vote all three up please.

cczech

Voted...

And also tossed a couple ideas.

Check out my ER - http://thwack.solarwinds.com/ideas/2556

adatole

Reminder, you can do it now with custom properties (as Shail alludeded to). Described here:

rob.hock

Check out the 10.7 RC with node and interface thresholds. NPM 10.7 RC1 Now Available

Yes, we're still working on volumes.

FormerMember

Hi rob.hock

Is it looking like volume thresholds likely to hit 10.7 GA? We are using the thresholds for CPU, Memory, packet loss and response time and it's looking very good and we would very to use these for setting up new volume alerts without the need to create custom properties.

Jon.

Update: I see NPM 10.7 has gone GA today.