[Linux-HA] pingd, quorum, split-brain... should I give up?
dejanmm at fastmail.fm
Mon Oct 22 12:05:58 MDT 2007
On Mon, Oct 22, 2007 at 02:45:42PM +0200, Riccardo Perni wrote:
> Hello to all,
> it is several days that I'm trying to set-up a split-site cluster but
> with scarce results.
> Since the two cluster nodes will be several Km away I cannot set up a
> reliable communication media between them, so I have to run heartbeat on the
> main Ethernet; I hoped that using pingd and an external ping site could
> help me to solve the potential conflict that will showup, but probably I'm
> not smart enough to solve this problem... Can someone help me?
> Actually I've set up a test using virtual machines
> Only one resource is running (a virtual IP using ocf:IPaddr) and only one
> constraint copied from linux-ha.org pingd FAQ.
> All seems to work right and if I broke the network connectivity of one of
> the nodes the resource is runned by the node with woking network; but both
> nodes get the "dc" status and when the connectivity is restored I have a
> split-brain condition with both nodes running the resource.
Right, because you can't prevent split-brain this way. What you
may prevent though is running a resource on the node which lost
connectivity. What does exactly happen once the connectivity's
restored? Can you post logs for that?
> Is it possible
> to handle this situation?
You may try quorumd. See
> Linux-HA mailing list
> Linux-HA at lists.linux-ha.org
> See also: http://linux-ha.org/ReportingProblems
More information about the Linux-HA