Lars Marowsky-Bree napsal(a): > On 2014-01-06T11:03:02, Jan Friesse <jfriesse@xxxxxxxxxx> wrote: > >> Change network.target to network-online.target to make sure network is >> really up and running when starting corosync. > > Is this sufficient? AFAIK it should be, but I'm really not .*d expert. The problem is really not in missing eth interface itself, but in missing lo interface. > > Wouldn't implementing the last section from > http://www.freedesktop.org/wiki/Software/systemd/NetworkTarget/ be > better? (Or perhaps this is already planned.) > > rtnetlink seems perfect, but a bit of trouble. But for ucast, wouldn't > IP_FREEBIND already be possible? I don't think it's needed. Corosync detect existence of interface by itself. So functionality already exists, with exception of lo (and yes, I'm planing to remove even this dependency, at least because many poeple does testing of HA by iptables -j DROP (what of course blocks also lo)). > > > Regards, > Lars > Honza _______________________________________________ discuss mailing list discuss@xxxxxxxxxxxx http://lists.corosync.org/mailman/listinfo/discuss
