On Sun, Apr 08, 2007 at 08:14:03PM +1100, Russell Coker wrote: > On Sunday 08 April 2007 16:46, Simon Horman <[EMAIL PROTECTED]> wrote: > > This seems to be a bit of an easy trap to fall into. > > Are there any fixes floating around? I was thinking > > that perhaps a cluster id of some sort would be a good > > idea. But I'm not sure. > > There is a cluster ID stored in the CIB. However that is going to be > copied if you copy both nodes including configuration. > > The ha.cf file already lists all nodes that are in the cluster via the > "node" directive. Surely if a node calling itself "foo" asks to join > the cluster then regardless of whether it has a suitable auth key it > should not be accepted if the list of valid nodes includes no "foo". > > Even if you have the case of a valid node in the cluster having the > wrong name due to a configuration error you can't keep a valid > configuration if you allow it to join as it makes the process of > determining quorum difficult. It's impossible to know whether it's a > backup copy of a node or a mis-named node. Allowing a machine with > the wrong name to join and then rejecting a machine with the right > name because the number of nodes specified in the config file have > already joined (as is currently the case) is just wrong.
Yes, I agree that sounds a bit silly. I'm actually surprised that what you describe is going on. Hopefully someone on the linux-ha-dev list can explain in a little detail what is supposed to occur in this kind of situation, and we can take things from there. -- Horms H: http://www.vergenet.net/~horms/ W: http://www.valinux.co.jp/en/ -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]