As soon as I changed all the systems gmond.conf and restarted the
daemons, every system was recognized.
It's one of those "But I want to know _WHY_" situations. When I posted
this back on 6/14 nobody responded with a why, but a some asked me to
post a reason if I ever find out.
Can anyone here explain why the mcast_channel value makes a difference?
I went through the same. Never figured out why, but I always assumed
that the network engineers and/or the specific switch gear involved was
the reason. But yes, changing the multicast address, in my case to
224.0.1.20 (used for "any private experiment" per RFC 1700) did the trick.
If I had free time, I'd run tcpdump on every node - before and after
changing the mcast_channel from the default...
-Matt Cuttler
