Greetings,
I've a small cluster of nodes that I'd like to see running gmond. I've
compiled, installed and fired up the gmond client daemon but something
about my configuration isn't right.
There are 6 clusters across 4 switches and 4 cabinets.
I think the relevant portions of the clients gmond.conf are:
cluster {
name = "cluster-1" # or cluster-2 ... cluster-6
}
udp_send_channel {
mcast_join = 239.x.y.z
port = 8649
}
where x.y.z is a local working multicast address.
In this configuration, the host running the web frontend and gmetad
presents the cluster as anything on the same switch (based on physical
connections rather than the logical "cluster-A".)
On a whim, and this is probably where my lack of understanding comes in,
I set data_source in /etc/gmetad.conf to something like:
data_source "cluster-1" 10.0.0.1 10.0.0.2 ... 10.0.0.N
data_source "cluster-2" 10.0.1.1 10.0.1.2 ... 10.0.1.N
...
data_source "cluster-6" 10.0.5.1 10.0.5.2 ... 10.0.5.N
Where 10.0.x.y are hosts in the cluster. I also changed gmond.conf on
the clients to use:
udp_send_channel {
host = "my-gmetad-host.f.q.d.n"
port = 8649
}
and commented out the mulitcast channel as I have it above.
In this configuration the gmetad server doesn't really do much, in-fact
all the clusters show no hosts down.
So, how what should I be doing to align clusters with their logical
names rather than by switch boundaries?
Cheers,
Ryan
