Carlo Marcelo Arenas Belon wrote:
> On Sat, Feb 09, 2008 at 11:24:01PM -0800, Ben Rockwood wrote:
>
>> I apologies for my ignorance, but could you please clarify how that is
>> accomplished? I'm not sure what you mean by "same unicast address".
>>
>
> select 1 node in your cluster (better if not memory bound) or even better the
> server which you are going to use as the gmetad/frontend server (let assume it
> is 10.10.0.20) and then set the following in /etc/gmond.conf in all nodes :
>
> udp_send_channel {
> port = 8649
> host = 10.10.0.20
> }
>
> remove any other udp_send_channel, or udp_recv_channel that reference
> multicast (because you are not using it) and be sure that you have in the
> collector the configuration to accept unicast messages (udp) and gmetad :
>
> udcp_recv_channel {
> port = 8649
> }
>
> tcp_accept_channel {
> port = 8649
> }
>
> if you did nothing else and everything is running by default you should be
> able to see your cluster through the web interface.
>
This worked beautifully! I previously had believed that each node
needed to be aware of every other node in some fashion and my
configurations were absolutely out of control. This method works
wonderfully.
I would really appreciate it if you could help educate me a bit on why
this works. Based on observation it would seem that gmond's send out to
the udp_send_channel some form of update which the "collector" gmond
then stores in memory for later requests. Is this roughly correct? My
mistake had been in believing that gmond's were entirely passive and
only returned data when explicitly requested.
Thank you very much Carlo!!!
benr.
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Ganglia-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ganglia-general
