Terry,
There are many ways to set this up, and I have a feeling that the way I
am going to recommend is not the way you will ultimately want to set
your ganglia up, but I am pretty sure it will work for your immediate goal.
In P.gmetad.conf:
data_source "P" 30 node1.P.com node2.P.com
gridname "PGRID"
authority "http://webfrontend.S.com/ganglia/";
In S.gmetad.conf:
data_source "S" 30 node1.S.com node2.S.com
# here we are polling the P's gmetad XML port, not a gmond port
data_source "P" 30 gmetadnode.P.com:8651
authority "http://webfrontend.S.com/ganglia/";
gridname "SGRID"
I would NOT recommend this config, but without knowing your network
situation (you mentioned private networks) or seeing your configs, this
is the only way I can recommend that I know should work. Ideally, I
would recommend having one webfrontend and one gmetad polling both
clusters; it would be much simpler and less prone to breakage.
Ian
Terry Gliedt wrote:
I have two clusters (S and P) each with it's own private network of
nodes. Both S and P run gmond and gmetad and I can see the data
collected for each in /data/gmetad/rrds. I've configured a web
interface for both which works, albeit with two URLs
Now it's time to mash these two together so I have one web application
(URL) showing the grid and I can select which ever cluster I'm
interested in.
I've tried endless combinations. My most recent attempt put two
'data_source' statements in gmetad.conf on S and I started getting data
in /data/gmetad/rrds for each cluster. "This must be it", I exclaimed!!
Alas, nothing shows up in the web interface. Can anyone explain the
necessary magic here? TIA
