Re: [Ganglia-developers] questions on metrics aspython modules

[Christian G. <chg...@wa...>]

No problem to give a "user oriented" view of what we do with "spoofing".
The clusters we manage in my compagny ( Alcatel-Lucent) are equipped with 24 
nodes, only two of the nodes are "main node" working in actif/standby, then 
only one of these two "main node" are operational at a time.
Only "main node" run gmetad, but all the nodes run gmond.
This "main node" already collect metrics from other nodes using a historic 
proprietary software and offer all these metrics through CLI, then all the 
metrics of all the nodes are already available on the "main node".
Then to bring all the user metrics of all the nodes in Ganglia we only have 
to deliver one python script on the "main node" collecting the metrics on 
the proprietary CLI of the "main node" at send it using gmetric with the 
"spoof" option to preserve in the front-end the real node origin of each 
user metric.
Without this option, we would need to deploy scripting on each node to 
collect locally the metrics...end also improve our proprietary software to 
open the CLI on no "main node"...moreover we only have python on the "main 
node"...
As a summary :  Without "spoofing" may be we would not use gmetric, and may 
be without gmetric we would not use Ganglia ?
Christian.


----- Original Message ----- 
From: "Brad Nicholes" <bni...@no...>
To: "Alex Dean" <al...@cr...>; 
<gan...@li...>; "Christian Gouret" 
<chg...@wa...>
Sent: Tuesday, November 06, 2007 7:36 PM
Subject: Re: [Ganglia-developers] questions on metrics aspython modules


>>> On 11/6/2007 at 11:26 AM, in message 
>>> <002401c820a2$9500d670$0c01a8c0@PC2004>,
"Christian Gouret" <chg...@wa...> wrote:
> Hello Brad,
> I read the README pointed by the link you provide, and I see nothing
> regarding the "spoofing" option introduced in the version 3.0.4 of 
> Ganglia.
> Does it mean that with "python module", it'll not be any more possible to
> use spoofing ( I can't believe it ) ?
> Christian.
>

That is actually a question that I have.  Up until now I have not touched 
the "spoofing" code at all.  What was there before, should still work in the 
code that is currently in trunk.  However, the XDR data packet restructuring 
that I am currently working on, affects the "spoofing" packets 
significantly.  To this point, I have not be able to completely understand 
what the "spoofing" code is doing.  I would like to understand the 
"spoofing" option better from a user point of view so that it can be either 
carried forward in 3.1 as-is or replicated in a python or C interface 
module.  Maybe you can enlighten me on what "spoofing" is, how it is used 
and why.  What is the expected out come or functionality?

Brad