Am 2012-06-18 21:33, schrieb Arnold Krille: > This is most probably not a problem of sheepdog but has to do with > corosync. > Normal corosync needs quorum to run, that is more then half of the > nodes need > to be up. On a two-node cluster this leads to the effect you see. > Either add a third node or configure corosync to ignore the quorum. > (But don't > add a third node _and_ ignore quorum, that opens a can of worms named > split- > brain.) Sorry for bring back this old mail... While testing with my current problem, i had a setup with four zones (on two host...). If I separated two of them (kill or plug network cable) the two remaining nodes try to recover, but failing (somehow expectable ;-). But with your description I await, that the remaining sheeps stops because two are not more than the half of all nodes... Cheers Bastian |