[Nagiosplug-devel] Suggested alterations to the Performance Protocoll

Ben Clewett Ben at clewett.org.uk
Tue Sep 14 07:09:07 CEST 2004


Hi Yves,

Thanks for the info.

If we use strings, my idea is more advise to the developer of the 
plugin:  Strings are legal.  These will be interpreted as a collection 
of compatible objects, or an enumerated list.  Such as can be displayed 
in a pie chart.

This will therefore allow the use of, say, IP addresses.  But not, say, 
arbitrary lines from a log file.

If the strings cannot be understood in a mathematical or statistical 
context, I think they should be added to the XML data, and not as 
performance data.  This gives the XML a valid use, and the performance 
data a valid use.

By enforcing this rule at this stage, we can give the user data of 
better quality which can be stored and analysed in a more effective way.

We have seen how unfortunately users will happily invent anything they 
can fit into a protocol and hope we can understand it later on.  This 
may just give the users a little help. :)

Ben



Yves Mettier wrote:

>>Yves,
>>
>>
>>>I suggest a new vote. Who is for/against:
>>>1| label=valueUOM;range... | <xml>extra data</xml>
>>>2| label=valueUOM;range... label='string'
>>>3| label=valueUOM;range... label='string' | <xml>extra data</xml>
>>>4| other suggestion ?
>>
>>Can I make a suggestion?  IF we include textual data:  Can we limit the
>>data strictly to comparable enumeration lists?  Eg, IP addresses.
>>Therefore stuff which can be drawn on a pie chart, gant chart, or
>>similar graphical output?
> 
> 
> I don't think we can limit that.
> For pie|gant charts, you can do it with any string.
> 
> 
>>Since I am the gue who will be programming these outputs, (with what
>>ever file name extension :), I would like to suggest this so that
>>something meaningful can be done with the data.
> 
> 
> select count(*) from table where count(value) > 10%
> With that, you forget strings that appear only once and draw data when they appear at
> least in 10% of the data. With the missing data, you can put "unknown" :)
> 
> Yves (SQL's worst programmer :)
> 
> 





More information about the Devel mailing list