After solving my /etc/hosts issue (in the host) by making it immutable, I still couldn't create a cluster. After a bit I found out hat it was due to Cluster.pm, so I tried to "fix" it like this:
root@hydra:/usr/share/perl5/PVE# diff Cluster.pm.bak Cluster.pm
182c182,183
< my $localip = $ifaces->{vmbr0}->{address} || $ifaces->{eth0}->{address};
---
> #my $localip = $ifaces->{vmbr0}->{address} || $ifaces->{eth0}->{address};
> my $localip = $ifaces->{eth0}->{address} || $ifaces->{vmbr0}->{address};
Before $localip would be vmbr0 by default, and now it is eth0. Since eth0 is normally a real network IP, but vmbr0 doesn't have to be, as in my case, this makes a lot more sense. It seems to be working; I even migrated (offline) a VE from the master to a slave. But without looking at the whole code in detail, I cannot tell if it will make trouble later along the way. I would like to hear the opinion of one of the developers if possible. Maybe someone knows why vmbr0 is normally the default IP?
root@hydra:/usr/share/perl5/PVE# diff Cluster.pm.bak Cluster.pm
182c182,183
< my $localip = $ifaces->{vmbr0}->{address} || $ifaces->{eth0}->{address};
---
> #my $localip = $ifaces->{vmbr0}->{address} || $ifaces->{eth0}->{address};
> my $localip = $ifaces->{eth0}->{address} || $ifaces->{vmbr0}->{address};
Before $localip would be vmbr0 by default, and now it is eth0. Since eth0 is normally a real network IP, but vmbr0 doesn't have to be, as in my case, this makes a lot more sense. It seems to be working; I even migrated (offline) a VE from the master to a slave. But without looking at the whole code in detail, I cannot tell if it will make trouble later along the way. I would like to hear the opinion of one of the developers if possible. Maybe someone knows why vmbr0 is normally the default IP?