Rob Murray

Moved back to draft for now ... the change will need to be rebased once the SetKey prestart hook is re-removed.

Rebased, and added a commit to re-remove the SetKey hook. It'll need to be rebased again once https://github.com/moby/moby/pull/47621 is merged (and that'll make `TestSetInterfaceSysctl` pass).

> 1. Keep track of children and only delete the interface when no more children exist > 2. Don't delete the dummy(/vlan) parent interface, but instead clean it up in...

> I was thinking about this a bunch and I think for vlan parents we'd need 1 either way, so probably I was focusing on the dummy interfaces too much....

If the internal nameserver kept hold of IPv6 upstream addresses (as well as IPv4 ones), I think it'd solve the problem. I should be able to have a go at...

Updated to reflect a last-minute change of heart on the sysctl naming (`net.ipv6.conf.eth0.disable_ipv6` was shortened to `ipv6.conf.disable_ipv6`, now it's `net.ipv6.conf.IFNAME.disable_ipv6`).

This change breaks the timeouts, I can raise a PR to fix that. But, the `EINTR` we've been seeing in moby CI is unrelated - it's due to https://github.com/vishvananda/netlink/pull/925, exposing...

HI @s4ke - thank you for the debugging, description and proposal. As I've not looked at Swarm in any detail yet, I think @corhere is much better placed to take...

> Once case which illustrates the need is where IPv4 is NATed but IPv6 is not. Since we won't be allowing port translation when NAT is disabled, `-p 80:8080/tcp` would...

Closing in favour of https://github.com/moby/moby/pull/47871 ... which disables NAT as well as masquerading.