Ganglia developers,
I believe I am using Ganglia in a similar fashion to Mike who has had
several config file postings the last few days. I may have gone about
things wrong in my config files but I would like to have a better
understanding so I will explain my configuration.
I have one gmetad node. This node runs on the host that my websever runs
on. I have other hosts that report data some on the same subnet as this
head node and some on other subnets.
I was able to do the logical groupings of my nodes mainly by using the
cluster names and then by changing the mcast_join address in the
udp_recv_channel and udp_send_channel. I used the default port.
On a few nodes I had to add a ttl = 2 or ttl = 3 parameter in some of the
the udp_send_channel information before I could see data from the nodes.
On the whole things are working like I wanted. However, I do have a few
problems which makes me feel as if I am not fully understanding how nodes
should be set up.
1. When I named the real "cluster" name in the gmond conf file for head
nodes data I totally lost my interface to ganglia. Given that I put
"unspecified" back into the gmond.conf file on the head node.
2. It is my understanding that the gmond on the head node receives the
data from the other hosts so I have udp_recv_channels with the various ip's
I set up my other cluster groups on in my head node's gmond.conf. See
clips of head node files below:
gmetad:
data_source "my cluster" localhost
data_source "mx hosts" mx2:8649
data_source "dns hosts" cuns3:8649
data_source "mevs HA hosts" lewis:8649
data_source "mail hosts" mail:8649
gmond.conf on head ganglia node:
cluster {
name = "unspecified"
}
udp_send_channel {
mcast_join = 239.2.11.74
port = 8649
}
/* You can specify as many udp_recv_channels as you like as well. */
udp_recv_channel {
mcast_join = 239.2.11.71
port = 8649
bind = 239.2.11.71
}
udp_recv_channel {
mcast_join = 239.2.11.72
port = 8649
bind = 239.2.11.72
}
udp_recv_channel {
mcast_join = 239.2.11.73
port = 8649
bind = 239.2.11.73
}
udp_recv_channel {
mcast_join = 239.2.11.74
port = 8649
bind = 239.2.11.74
}
The problem I am having is that even though my nodes show up in the cluster
specified in their gmond.conf file several show up in this "unspecified"
cluster. All of the nodes from the "mail hosts" cluster show up they are
communicating on the 11.74 ip and have the ttl = 2 within their
udp_send_channel setup and the other nodes that show up also have the ttl =
2 parm.
If someone could shed some light on these problems and if I approached this
wrong I would greatly appreciate it.
Thanks,
Dierdre
