i am running a BM 3.8.4 ir1 clustered configuration on two NetWare 6.5.4 servers at multiple clients. generally it is working well, but over time i have discovered an annoying problem. if i make changes to the configuration of the cache or anything directly related to the proxy, when i click ok to save changes, proxy reinitializes. generally this is a good thing, but it does something that messes up the arp table related to these cluster resources. i get a duplicate ip warning even if the cluster resources aren't currently on that node. if i watch proxy stats, i will start seeing user connections climbing on the node that just reinitialized proxy.nlm even though the ip address are bound to the opposite node. it seems to me that some sort of arp gets sent out and the switch now thinks that the ip is associated with the wrong node.

it does some screwy things when this happens and i generally migrate the resource so that the ip matches the arp that just happened.

is there a way to prevent this or make it work differently? manually setting "allow duplicate ip addresses=on" before making a change and then flushing the arp table on the server after this happens seems to help, but i haven't had enough time to be sure.