The dreaded hang-on-install after dhcp v6 configuration

I think this has been discussed elsewhere. I remember seeing this and posting before about it, I think at the time I just dropped the idea of using OMV because I couldn't be bothered and installed FreeNAS. Now I decided to try a bit harder.

Install into VirtualBox with nat network works fine.

Install into VirtualBox with bridged network does not work.

I remember when I was installing into Proxmox it wouldn't work either, but I only ever create VMs in Proxmox with bridged networking.

So there is something 'shitty' about my network, or at least there's something upsetting OMV, and it must relate to ipv6. Or so I assume.

I tried installing without configuring the network (cable unplugged), but my first attempt at this left me with no networking on boot. I wondered if I should try to 'fix' my network instead.

So I took a look at my broadband router. I've got an arrangement where I've disabled the DHCP on the router, and I run Dnsmasq for this. So the router is only ever doing routing, I control my DHPC/DNS outside of that. I don't see any ipv6 settings for DNSMasq that would make any difference here.

So I thought I would post to see if anyone has any ideas. Are there specific settings I'm looking for either on Dnsmasq or my router that I can play around with for instance?

Thanks!

Yes, that's my last resort. I've not disabled DHCP completely, just on my DSL router, I've still got a proxmox VM with DHCP/DNS on it on the same network. I assign the IP address (which is static) via a specific MAC and that's usually how I setup most home machines.

I've now tried killing the dhclient process (it forks so there are two PIDs to kill, no killall in th alt-f2 shell), and that does seem to allow the install to complete and networking is OK. Just tested that in VirtualBox Bridged. So that's my fallback. I'm mystified why there's a problem though. I added the line

dhcp-range=2001:db8::02, 2001:db8::ff,12h

to my /etc/dnsmasq.conf hoping it might serve some addresses.

tcpdump shows some ipv6 activity from the MAC of the OMV VM but I'm not an expert on it so it doesn't make much sense to me.

There was this:

ICMP6, neighbor solicitation, length 32, who has fe80::a00:2eff:ff2e:7f9f
      unknown option (14), length 8 (1): 
        0x0000:  2276 3220 c2aa

Going to try to investigate a bit more....

Maybe I didn't express myself very clearly there. I mean I'd like to understand the problem. Who knows, maybe contribute a fix? It's not like I want to spend a week on it, but I have some free time tonight.

I am using DHCP. I think the problem is it's serving an ipv4 address, but it's not serving correctly the ipv6.

For others that may encounter this issue, it appears this is the result of a single setting in my router. I had enabled ipv6, but disabled DHCPv6.

It's not the overall ipv6 enabling but I believe the RA setting just below it which would have been activated.

You can read about the RA modes here: https://docs.netgate.com/pfsen…ervices/dhcp/ipv6-ra.html

Once the ipv6 setting was off (or conversely the dhcpv6 server switched on) everything worked.

I think that the presence of RA signals the Debian net cfg code to persistently try to get an address, and that may be a bug in that it just doesn't support the so-called 'router only' mode, but I've more reading to do about ipv6 before I can be sure.

And finally, if you want to reproduce the problem you can do it in VirtualBox. Simply setup pfsense (I used version 2.5.2 CE) with a private network one side and your normal LAN the other. Create an OMV virtual machine on the private network. You don't even need to configure anything, out of the box pfsense will give the private network an environment where OMV install fails. Once the Router mode is set to disabled, or the dhcpv6 server is started, the install succeeds.