Hi list.
I recently ran into a situation where a local router decided to cut off my
lab's network from the network where our DNS servers are located, which
resulted in the loss of name resolution services. This, as far as I can tell,
resulted in the shut-down of my 5-node cluster.
The cluster is configured with two broadcast interfaces: one on a local,
non-routable network (192.168.x.x) and one on a valid internet network
(16.x.x.x -- one of the original Compaq class A nets). My ha.cf is configured
to use bcast first on the local net, and then on the routable net. The
interface for the local net is listed first so that it's used first. The DNS
server is on the routable interface.
My understanding with heartbeat is that the bcast directive tells it to use UDP
broadcast on the interfaces in the order in which they're given. UDP broadcast
doesn't rely on DNS, so I'm wondering why a loss of DNS resolution would cause
the cluster nodes to not be able to hear each other?
Thanks!
- Conor
PS: I'm happy to send logs, but I didn't think they'd be necessary for this
question...
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
