Ethan Russell | 21 Jul 18:13 2010
Picon

[Check_mk (english)] Question on CPU load check

Hi all,

I'm running into a strange issue with the CPU load check, namely the way the check seems to be working is that my multi-CPU machines are reporting back performance data as a server with only 1 CPU would, but it is multiplying the warning & critical levels by the number of CPUs!

Currently here's my config:

Server is running Check_mk version 1.1.6 on CentOS 5 x64

In main.mk I've set the CPU load defaults levels as:

cpuload_default_levels = ( 4.0, 8.0 )

Checking one of my single CPU servers, all seems well:

check_mk -vpn Syslog
Check_mk version 1.1.6
CPU load             OK - 15min Load 0.84 at 1 CPUs                           (load1=0.64;4.0;8.0;0; load5=0.78;4.0;8.0;0; load15=0.84;4.0;8.0;0;)

However, when I check my DB server the warning levels are multiplied but the actual load data is returned as a server with a single CPU would but multiplies the warning & crit levels:

check_mk -vpn DB0
Check_mk version 1.1.6
CPU load             OK - 15min Load 0.00 at 8 CPUs                           (load1=0.05;32.0;64.0;0; load5=0.03;32.0;64.0;0; load15=0.0;32.0;64.0;0;)

End result?  My database servers with 8 CPUs in them never fire off notifications when under heavy load, because they think the warning and critical levels are eight times higher than they actually are.  Is this intended behavior? If so, what would be the recommended method for me to correct this?

Thanks!

ER




_______________________________________________
checkmk-en mailing list
checkmk-en@...
http://lists.mathias-kettner.de/mailman/listinfo/checkmk-en
Mathias Kettner | 27 Jul 09:00 2010
Picon

Re: [Check_mk (english)] Question on CPU load check

Am 21.07.2010 18:13, schrieb Ethan Russell:
> Hi all,
>
> I'm running into a strange issue with the CPU load check, namely the way
> the check seems to be working is that my multi-CPU machines are
> reporting back performance data as a server with only 1 CPU would, but
> it is multiplying the warning & critical levels by the number of CPUs!
Thats by design. The levels are *per CPU*. Please see
check_mk -M cpu.loads for details.

"CHECK PARAMETERS 

  warning  (int): 15  minute average  load per processor  at which  a 
warning is
  triggered. If you set this value to 2 on a 16 CPU machine, then a 
warning will
  be reached at a load of 32.        "

>
> Currently here's my config:
>
> Server is running Check_mk version 1.1.6 on CentOS 5 x64
>
> In main.mk <http://main.mk> I've set the CPU load defaults levels as:
>
> cpuload_default_levels = ( 4.0, 8.0 )
>
> Checking one of my single CPU servers, all seems well:
>
> check_mk -vpn Syslog
> Check_mk version 1.1.6
> CPU load             OK - 15min Load 0.84 at 1
> CPUs                           (load1=0.64;4.0;8.0;0;
> load5=0.78;4.0;8.0;0; load15=0.84;4.0;8.0;0;)
>
> However, when I check my DB server the warning levels are multiplied but
> the actual load data is returned as a server with a single CPU would but
> multiplies the warning & crit levels:
>
> check_mk -vpn DB0
> Check_mk version 1.1.6
> CPU load             OK - 15min Load 0.00 at 8
> CPUs                           (load1=0.05;32.0;64.0;0;
> load5=0.03;32.0;64.0;0; load15=0.0;32.0;64.0;0;)
>
> End result?  My database servers with 8 CPUs in them never fire off
> notifications when under heavy load, because they think the warning and
> critical levels are eight times higher than they actually are.  Is this
> intended behavior? If so, what would be the recommended method for me to
> correct this?
>
> Thanks!
>
> ER
>
>
>
>
>
>
> _______________________________________________
> checkmk-en mailing list
> checkmk-en@...
> http://lists.mathias-kettner.de/mailman/listinfo/checkmk-en

Gmane