[Nagiosplug-devel] check_load, inaccurate results.

Ben Clewett Ben at clewett.org.uk
Mon Nov 15 06:57:02 CET 2004


Andreas,

Thanks for your honest answer.  I will see what can be done.  For 
instance replacing this plugin with the same plugin as published several 
months ago.  That version worked perfectly...

Ben



Andreas Ericsson wrote:

> Ben Clewett wrote:
> 
>> Dear plugin developers,
>>
> 
> You're wasting your breath. (Official) plugin development has been at a 
> standstill for a very long time. I suggest you take a stab at making 
> whatever changes are necessary yourself. If you spend 5 minutes a week 
> it will be faster than waiting for the plugin developers to take action.
> 
>> I have noticed that the check_load only reports integer performance 
>> data.  Eg:
>>
>> OK - load average: 0.62, 0.68, 0.58 | load1=0;1;2;0 load5=0;2;3;0 
>> load15=0;3;4;0
>>
>> This plugin used to return correct values.
>>
>> I note the new perfdata call in utils.c:
>>
>> char *perfdata (const char *label,
>>  long int val,
>>  const char *uom,
>>  int warnp,
>>  long int warn,
>>  int critp,
>>  long int crit,
>>  int minp,
>>  long int minv,
>>  int maxp,
>>  long int maxv)
>> {
>>
>>
>> Performance data from a few plugins is required to respect floating 
>> point numbers.   Can I suggest this function is changed to accept real 
>> numbers and not integers?
>>
> 
> It's rather stupid to have a unified perfdata function, since plugins 
> will return highly different kinds of values (as you have just pointed 
> out). It is very impractical to convert integer values to floating point 
> ones, and what's being accomplished can usually be done with a simple 
> printf(3) statement anyway. Where it cannot, I see no way for the plugin 
> to correctly utilize the perfdata function either, so it's existence is 
> fairly useless.
> 
>> I also note that this function call cannot handle the range values as 
>> defined in:
>>
>> http://nagiosplug.sourceforge.net/developer-guidelines.html#THRESHOLDFORMAT 
>>
>>
>> Which some developers may like to respect.
>>
>> Regards,
>>
>> Ben Clewett.
>>
> 





More information about the Devel mailing list