avg and cpu.load

[Tom Decaluwe]

Hi, I'm new to zabbix and trying to figure out how the triggers work on the CPU load. After activating the default windows template i'm getting alot of CPU.load emails. I know my systems are well used and i want to finetune the trigger to be less agressive and report me if the CPU load has been over 90% for the past 15 minutes.

the default trigger is: {Template_Windows:system.cpu.load[,avg1].last(0)}>5


now it seems i have a few thing i can play with but I need some help in understanding this.

1) the first parameter is: avg1, i can see there is also an avg5 and avg15. Am i correct in interprating this as:

- avg1 => the average of all values for the last 1 min
- avg5 => the average of all values for the last 5 min
- avg15 => the average of all values for the last 15 min

2) last(0) i'm guessting is the last value?

3) >5 => means the value is higher than 5%

Can i thus create a trigger as follows that would trigger if the average load is higher than 90% for the past 15 min:

{Template_Windows:system.cpu.load[,avg15].last(0)}>90


or would it need to be something like this

{Template_Windows:system.cpu.load[,avg1].avg(900)}>90

many thanks,

Tom

[elvar]

Hi, I'm new to zabbix and trying to figure out how the triggers work on the CPU load. After activating the default windows template i'm getting alot of CPU.load emails. I know my systems are well used and i want to finetune the trigger to be less agressive and report me if the CPU load has been over 90% for the past 15 minutes.

the default trigger is: {Template_Windows:system.cpu.load[,avg1].last(0)}>5


now it seems i have a few thing i can play with but I need some help in understanding this.

1) the first parameter is: avg1, i can see there is also an avg5 and avg15. Am i correct in interprating this as:

- avg1 => the average of all values for the last 1 min
- avg5 => the average of all values for the last 5 min
- avg15 => the average of all values for the last 15 min

2) last(0) i'm guessting is the last value?

3) >5 => means the value is higher than 5%

Can i thus create a trigger as follows that would trigger if the average load is higher than 90% for the past 15 min:

{Template_Windows:system.cpu.load[,avg15].last(0)}>90


or would it need to be something like this

{Template_Windows:system.cpu.load[,avg1].avg(900)}>90

many thanks,

Tom

You may want double check me on this but I'm pretty sure cpu.load is not the same as the CPU utilization % you see when viewing Task Manager. If you want to monitor the CPU utilization % you see in Task Manager, you need to tie into the performance counters. I know Zabbix supports the performance counters though I don't remember the exact syntax.

[Tom Decaluwe]

Thanks for the reply, i don't know what the difference is between the perf counter and the system.cpu.load but just to be safe i did configure a perf counter object and will use this as this better refelects the data i actually want to monitor.

I am however experiancing some issue with this item but posted in a dffirent thread.

Does anyone know the difference in avg, avg5 and avg15 as this is still very unclear to me and i see these values pop up quite ferquently.

Thanks,

Tom

[harmonica]

I Tom,

Maybe this help: http://en.wikipedia.org/wiki/Load_(computing)

[subba5678]

Hi Tom ,
Iam facing the exact problem in creating a trigger for CPU Utilization for "90%" . Could you please post me the macros for performance counters which i need use to create Item and Trigger for CPU utilization

Thanks in advance
Subbu

[aethos]

Never too late

I know this is an old topic but since google seems to find it first, here's an answer. First, for years unix systems have had load averages computed in real time, 5 minute and 15 minute averages. When you do a top you will see those counters.

The way to configure this (and to do it by % as well if that's the preference) is first to create a new item. Triggers just monitor the data returned from items. Unless you have an item with the data you can't write a trigger on it. The details on my item:
Key: system.cpu.load[,avg15]
Type of information: Float
Applications: Performance

The rest is mainly default. Once you have that item getting collected from your servers (note that you may have to activate it in your template and check that it comes up as active on your host) then you can create a trigger to alert if it goes over a certain threshold.

As for the value of load averages, the article states it clearly - the load average is the number of processes waiting for attention. A number between 1-2 means a fairly active system. If you get up in the 4-5 range or higher your system is in big trouble.

Hope this helps!

[HullZabbix]

For monitoring our Windows servers I use

{TemplateServers:system.cpu.util[,,avg1].avg(600)}>90

This triggers when the CPU (polled every 120 seconds in item config) utilisation, is over 90% for 10 minutes.