Bernard Li wrote:
I assume that you are using different mcast ports for the different data sources.
No, I've got them all on the same port, but the partitions can't see each other, so each of the data sources considers itself authoritative (if you will) for the cluster. I can try putting them on different ports, but it's not clear to me how that'll help.
Typically I find that if there are some errors with the metrics after I have done some modifications to the cluster (add/remove nodes) - I have to stop all the daemons (gmond/gmetad) and restart them. It usually takes some time before the data gets properly propagated and eventually the correct information shows up.
This has been fairly consistent for a few days (since I started using Ganglia), and doesn't seem to be affected by restarting daemons.
