gmond will query one data source for each cluster, and will only try another one if the first one fails. Membership in the cluster is determined by which nodes that gmond (the first data source) can see. Therefore, a cluster is a set of gmond nodes that see each others' metric traffic on the network.
If you're using multicast and all of your gmond nodes are in the same multicast domain, then they will all see each other if they're configured to use the same multicast IP and port. Any gmond that gmetad contacts, will provide information about all of the nodes, so they will all be one cluster. You can split them up by configuing some of them to use another multicast IP address, or to use a different UDP port on the same IP address. If your clusters are on physically separate networks (or VLANs) you may be able to keep them separate simply by ensuring that multicast traffic on the IP address they're using, isn't passed by the routers between those networks. If you're using unicast, then you directly configure which gmond nodes speak to which other gmond nodes, and build up your clusters one by one. -- Cos ------------------------------------------------------------------------------ Download Intel® Parallel Studio Eval Try the new software tools for yourself. Speed compiling, find bugs proactively, and fine-tune applications for parallel performance. See why Intel Parallel Studio got high marks during beta. http://p.sf.net/sfu/intel-sw-dev _______________________________________________ Ganglia-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/ganglia-general
