Hi all, It is my understanding that when setting up multiple clusters, you must have a different instance of gmond running as a 'head node' for each cluster (with multicast, every node can be head node, with unicast, it is target host). Part of the gmond configuration on a given node is to assign a 'cluster name' that is then used when displaying the cluster through the web interface.
My question - why not use that cluster name to determine cluster membership? (and are there any plans to do that in future versions?) I've been watching this list for a while, and the most common answer to questions having to do with trouble getting multiple clusters is to put each cluster on a separate port. If you're using multicast, it is pretty easy to point gmetad to multiple data sources (because any node can be a data source), but if you're using unicast, you either have to have separate head nodes or separate instances of gmond running on your aggregator host, both of which seem kind of annoying. The parameter to identify to which cluster a node belongs is already present in the cluster name parameter; why must we additionally corral it off using a different tcp/udp port? It seems to me that using the name to determine cluster membership would simplify things for the people configuring ganglia. Thanks, -ben -- Ben Hartshorne email: [EMAIL PROTECTED] http://ben.hartshorne.net
signature.asc
Description: Digital signature
