Closed kkremitzki closed 4 years ago
Yes, that's normal and you'll need to manually modify the network config of the container to account for that.
The short version is that short of running something like networkd or NetworkManager which can detect either the RDNSS field in the RA or detect the managed/other-config field in the RA and then do DHCPv6 to get the DNS, there's nothing to actually configure your DNS server.
The only reason why you have an IPv6 address at all is because address and route config from RA is automatically handled by the kernel, however the kernel doesn't care about DNS, causing what you're seeing.
ifupdown offers no way to do what networkd does and effectively detect what's supported through RA and go with that.
So you can definitely configure something like inet6 auto
with dhcp 1
or whatever the syntax is to get stateless DHCPv6 but if you then turn off IPv6, rather than just move on, your container will hang indefinitely, making such a config unsuitable for a default config.
For example, works on bionic+ since systemd-resolved is used, but is broken on xenial, buster+.
Reproduction setup:
Result:
Bionic:
Buster:
Xenial:
Apt failure: