Alarms - how is the sampling done and how often - Can we control it?

[Marc_Alumbaugh]

I have some alarms that occur on both VMs and Hosts.

The Alarms are High CPU and you can see the alarm definitions just fine..

The question is how often is the sampling being done and can I control the frequency of the sampling?

So here's an example..

The VM's normal CPU usage is say 10% or whatever.

something happens and it spikes to say 95% but only stays there for 15-20 secs.

An Alarm fires but I don't want it to fire since it's just a spike

So the questions are

What is the sampling window/speed of the statistics?

Is it every 20 secs?

Can I adjust the sampling speed look at a the average over a longer window?

Say it does sample every 20 seconds.

Can I say somehow to use the average of 3 samples? Before an alarm occurs?

[TiBoReR]

Same thing here.

I would like to not have an alarm when it is only a cpu or memory spike. But if it stays for 30 seconds, I would like to receive that alarm.

[shalpern]

It would appear that you have both come the same conclusion I have. You can not configure the alerts to be for x occurances for over x time. You should disable these alarms and use a 3rd party utility (e.g SCOM2007/NWorks) for monitoring.