CPU Monitoring problem

[RunningWave]

I've trying to monitor Fedora Core 4 server via ucd/net SNMP

The box is 2CPU Xeon x86_64 (with HypperThreading) which gives us 4 logical CPUs.

This seems to cause two problems with cacti:

- User CPU is reported as "na" I suspect this is because load is more than maximum which is set to 100

- System CPU is reported at 12% while I see it is 3% via vmstat which gives me the hint it is wrongly computed.

Proposal: I guess template need to be changed so maximum value is increased to 1600 or something, so it can work with more than once CPU. Alternative is to take number of CPUs into computation which might not be available via snmp.

[gttommy]

Hi, are you also asking for help? If you are, you can use a script to run something like "vmstat 2 2 | tail -1" (delay required because initial output won't be accurate), until you get SNMP to work for you.

[RunningWave]

Hi, are you also asking for help? If you are, you can use a script to run something like "vmstat 2 2 | tail -1" (delay required because initial output won't be accurate), until you get SNMP to work for you.

Thanks.

Well. If that is other host it would be more complicated. Of course I can do it via SSH but it is kind of slow and ugly.

I mainly looking to hear if someone can confirm this problem and if there is known work around for it

[Burnout]

Please take a look @ my topic, it handles the same problem. http://forums.cacti.net/viewtopic.php?t=10761

[RunningWave]

Thanks,

I personally just hacked templates so they allow up to 400 maximum.
It is better for me than creating different templates for different hosts.

I just think default template should be fixed this way so it does not breaks on dual CPUs and HTs

Showing over 100% of CPU usage is better than showing nan