On Tue, 6 Nov 2007, Brad Nicholes wrote:

> >>> On 11/6/2007 at 1:50 PM, in message <[EMAIL PROTECTED]>, "Brad
> Nicholes" <[EMAIL PROTECTED]> wrote:
> >>>> On 11/6/2007 at 1:45 PM, in message
> > <[EMAIL PROTECTED]>, Matthias
> > Blankenhaus <[EMAIL PROTECTED]> wrote:
> > 
> >> 
> >> On Tue, 6 Nov 2007, john allspaw wrote:
> >> 
> >>> Hey all - 
> >>> 
> >>> While I love the idea of the python module, I'd like the plain-old 
> >>> gmetric 
> >> binary to stay as-is. I'm not sure if there are/were plans on getting rid 
> >> of 
> >> it in favor of the new custom framework stuff. Some of the reasons why I 
> >> think it should stay:
> >>> 
> >>> - Gmetric makes building custom metrics insanely easy. I tell people all 
> >>> the 
> >> time: "if you can make a number with any script you have, you can get it 
> > into 
> >> ganglia, with whatever language you want." 
> >>> 
> >>> - We have a lot of custom metrics that depend on other software, (we take 
> >> results from nagios checks and spit them into ganglia with the gmetric 
> >> command, for example) and that can't be rewritten without a lot of work.
> >>> 
> >>> A huge part about why some folks where I work (Yahoo!) are interested in 
> >> ganglia is that even with as basic install, you get a LOT
> >>> of visibility into your systems and network, with very little effort. 
> >>> Being 
> >> able to embed a one-line gmetric command into any other monitoring they 
> >> already have is the icing on the cake.
> >>> 
> >>> Thoughts ?
> >> 
> >> I also vote for keeping gmetric in the game.  I have asked this question 
> >> before and I now sense a chance to reissue it :-).  Here is goes:
> >> 
> >> Is there general interest in extending gmetric so that it can take a list 
> >> of metrics, instead of just one ?
> >> 
> >> We have a situation where we need to feed in lots of metrics and thus
> >> suffer from the resulting system calls.
> >> 
> >> Another thought I had to further reduce system latency was to daemonize 
> >> gmetric and let it read from a local socket or pipe or whatever peopel 
> >> think is most convinient.
> >> 
> >> Cheers,
> >> Matthias
> >> 
> >> 
> > 
> > The good news is that the gmond module interface (python or C) takes care 
> > of 
> > those issues.  This isn't a reason to get rid of gmetric, just a reason to 
> > consider moving your metrics to the module interface rather than enhancing 
> > gmetric.
> > 
> 
> Another thought along these lines is that it might be easier to build a Gmond 
> module that will execute a given list of scripts and return the results as 
> individual metrics rather than enhance gmetric.  By implementing a gmond 
> script execution module you get daemonization for free as well as metric 
> scheduling.  Gmond already provides us with the infrastructure we need to 
> collect metrics and modules allow us to extend it in any way that is 
> necessary.
> 

Cool.  Are there examples for this available or documention ?

I assume this Gmond module would have to be written in C and is loaded as 
a DSO into gmond ?


Thanx,
Matthias

> Brad 
> 
> 
> 

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Ganglia-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ganglia-developers

Reply via email to