On Wed, Mar 05, 2008 at 12:04:56AM +0200, Ido Levy wrote: > > Hello, > > I am new to Ganglia and try to find some documentation in order to better > understand how should I configure a simple cluster.
one good description of the data flow is available from the following image: http://www-941.ibm.com/collaboration/wiki/download/attachments/3181/ganglia_cluster.gif the gmond and gmetad man pages are also a really good source of information. the default configuration are usually enough to get you started if you run gmond in all nodes (including the one where you are running gmetad). > I have three machines. Two of them are running gmond and the third one is > running gmetad. I am trying to figure out what parameters > should be configured on each side ( gmond and gmetad ) a cluster of 2 nodes (your gmetad is not part of the cluster but just used for monitoring) > I used the following but it didn't worked: > > /etc/gmond.conf > > cluster {. > name = "LinuxCL" > owner = "unspecified" > latlong = "unspecified" > url = "unspecified" > } > > /* The host section describes attributes of the host, like the > location */ > host { > location = "unspecified" > } > > /* Feel free to specify as many udp_send_channels as you like. Gmond > used to only support having a single channel */ > udp_send_channel { > host = <HOSTNAME_OF_GMETAD_MACHINE> > port = 8649 > ttl = 1 > } ttl = 1 is usually needed for multicast setups instructing the switches that the packets generated should be dropped in the gateway and not routed. when you use "host" for your "udp_send_channel" then you are using unicast for your setup and therefore "ttl" is most likely not what you want. > /* 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. > } this means you will be listening in a multicast channel, and not what you would need if using unicast as indicated above. for a unicast collector you want to have instead : udp_recv_channel { port = 8649 } > /etc/gmetad.conf. > > data_source "LinuxCL" localhost your head node (or the node where you run gmetad) should have a gmond running as a collector (based on your configuration above) which is going to be the one that gmetad pokes for the cluster information using tcp and therefore should have a configuration like : tcp_accept_channel { port = 8649 timeout = -1 } Carlo ------------------------------------------------------------------------- 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
