Re: Monitoring the wrong thing?

[Martin Pala]

The configuration is OK, but the limits should be modified. It seems 
that your system has CPU usage spikes which last for 2+ cycles and it 
trigger the alert. The limit should be set so, that you get alert only 
if the state is abnormal/pathological. What is normal/abnormal is 
specific for each system - you can watch the load and then set the 
limits accordingly ... for example rise the cpu(user) usage to 90% for 
10 cycles.

Regards,
Martin


On 08/04/2010 09:30 PM, Marc Pinnell wrote:
> Finally got Monit going this am on my webserver (daemon, 2 min interval). Since 
> then I am getting a couple of warnings an hour about high CPU loads. I am 
> monitoring the wrong thing (I don't totally understand the UNIX terminology 
> about loads)? Here is my config:
>
>    check system 1027mail
>      if loadavg (1min)>  4 then alert
>      if loadavg (5min)>  2 then alert
>      if memory usage>  75% then alert
>      if cpu usage (user)>  70% for 2 cycles then alert
>      if cpu usage (system)>  30% for 2 cycles then alert
>
> and a warning I just received:
>
>
> Begin forwarded message:
>
> > Resource limit matched Service 1027mail
> >
> >         Date:        Wed, 04 Aug 2010 15:24:13 -0400
> >         Action:      alert
> >         Host:        1027mail
> >         Description: cpu user usage of 80.8% matches resource limit [cpu user 
> > usage>70.0%]
>
> and then two minutes later:
>
> Resource limit succeeded Service 1027mail
>
>         Date:        Wed, 04 Aug 2010 15:26:19 -0400
>         Action:      alert
>         Host:        1027mail
>         Description: '1027mail' cpu user usage check succeeded [current cpu 
> user usage=0.0%]
>
>
> This is the way they all go so far. Seems like if it happens to check at the 
> very moment a web request comes in (which obviously happens on a regular 
> basis!), it trips the warning.
>
> Suggestions?
>
> Marc