Duplicate IP address after mysql stop

[Paolo Diana]

Version: mmm-mysql 2.0.10 on SLES11, compiled from source.
I have a two-nodes cluster, with a 'writer' exclusive role and a 'reader' balanced role with two IPs.
I noticed that, when mmm agent is down on a node, and I stop mysql on that node, monitor tries to move all roles on the other node, but as agent is down IPs are not stopped on the failed node, so I have duplicate IPs on the two nodes.
I understand I need a kill_host binary, but I think that if an IP is still pingable monitor dosn't have to activate it on an other node (I noted that monitor knew that the IP was up, but the agent was down, so it could know that activating IP on the other node was a mistake, and a very dangerous one).
Hope you can agree this is not a correct behaviour, and you can fix it.
Many thanks.

[Paolo Diana (diana-mondadori)]

Any news?

[Pascal Hofmann (pascalhofmann)]

I think it is correct behavior. I've had crashed hosts which still responded to ping fine. Use a kill_host binary.

[Paolo Diana (diana-mondadori)]

Thanks for you answer.
I really don't agree: I think it's not useful to activate the same ip on the second node, because I know it can't work well if it's a duplicate IP, and it's a risk anyway.
It should be better to have at least a configuration option to choose what to do in this case.
Thanks again.
Paolo.