Jesse Alvarez <[email protected]> wrote: > Good question. My understanding is that you need one gmond as a > collector per cluster. So let's say "red App cluster" has 100 nodes in > its cluster. One of those or a separate instance somewhere would have to > be designated the collector. Since only one collector can run on a host,
> Or am I totally missing the point of this? You seem to be using multicast, which is what I prefer. With multicast, every gmond is a "collector" without any extra effort on your part. Any gmond that can see multicasts from other gmond nodes on its cluster will collect all of their metrics, so if every cluster is configured to multicast to the same port & IP across the cluster, any one of those gmond's can be a data_source for your gmetad for that cluster. You can list multiple nodes on the data_source line for each cluster, so gmetad can fall back if the first one is down. As long as they're nodes that see the same multicasts, they'll have equivalent data. -- Cos ------------------------------------------------------------------------------ Check out the new SourceForge.net Marketplace. It is the best place to buy or sell services for just about anything Open Source. http://p.sf.net/sfu/Xq1LFB _______________________________________________ Ganglia-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/ganglia-general
