On 06/16/2011 04:28 AM, Jack Berg wrote: > > I have a two node cluster using heartbeat and haproxy. Unfortunately it is > impossible to provide redundant heartbeat paths between the two nodes at > different sites so it is possible for a failure to cause split brain. > > To evaluate the impact I tried disconnecting the two nodes and I found that > both become active and both try to keep the VIPs after the link is restored.
What do you mean by "disconnecting": what's your failure scenario and how do you expect it to be handled? Running daemons are not guaranteed (arguably, expected) to notice when the network cable is unplugged. You have to monitor the link and restart all processes that bind()/listen() on the interface. If your nodes are at different sites, you need to also deal with the loss of link at the switch, gateway, etc., and figure out which one is still connected to the Internet -- and gets to keep the VIP. Which in general can't be done from the nodes themselves. Dima -- Dimitri Maziuk Programmer/sysadmin BioMagResBank, UW-Madison -- http://www.bmrb.wisc.edu
signature.asc
Description: OpenPGP digital signature
_______________________________________________ Linux-HA mailing list Linux-HA@lists.linux-ha.org http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems