[collectd] Monitoring itself

[Luboš Staněk]

Hi Florian,
I am feeling better after reading that I was not a lunatic all the time. :)

Florian Forster napsal(a):
> Hi Lubos,
> 
> On Fri, Nov 10, 2006 at 05:33:26PM +0100, Lubo?? Stan??k wrote:
>> I will probably get the reputation of the permanent nagger. :)
> 
> No, I welcome new ideas :)
> 

Did you pass out some course about moderating OSS forums? :)


>> I have thought about monitoring the collectd server itself. The
>> interesting stats would be the time spent in the plugin's read and
>> write functions in the local mode or network load in the server mode.
> 
> Hehe, having statistics of an statistics daemon is kind of consistent ;)
> 

Well, it looks funny.
I do not want to get stuck in a loop. The stats of the daemon and
plugins are the last point, I do not want to gather stats of stats... :)


>> It could be configurable by the configure script in both non- and
>> debug build.
> 
> Counting the network-traffic and providing those counters to a plugin
> shouldn't be a problem at all. Measuring the time spent in _each_ plugin
> should explicitely be configured, since at least one extra call to
> `nanosleep' for each plugin is necessary. Measureing the overall
> `roudtrip time' shouldn't be a problem, since we need the time there
> anyway.
> 

I have thought about a plugin without read() that could be configured
and called as the last plugin. It would get stats from the global
variables provided by the daemon for local mode.
I have not analysed the server mode yet.


>> I will try to implement it for my own testing purposes. I want to
>> profile the ignore list with and without regex functions.
> 
> Okay, but please keep the patches separate; I much rather have a big
> number of small patches that add little functionality one by one that
> one huge patch that is hard to read and understand.
> 

I did only a hack for testing the regex list.
It is not in the form of usable patch.
I injected the plugin_read_all() with the check for p-type == "sensors"
| "sensor_voltage", DS definition, formatting the time spent and call to
rrd_update_file.


>> P.S.: I do not ask for help but opinions whether somebody would
>>       consider it useful.
> 
> Useful I don't know, but `consistent' and at the edge to `self ironic',
> definitely `yes' ;)
> 

Thanks for the sympathy... 8-)

> Regards,
> -octo

Best regards,
Lubos