Pulling containers is broken (dial tcp [IPv6]:443: connect: network is unreachable)

[MidnightSlicer]

Describe the bug
Docker changed the way their stuff works on the backend and it seems to have broke this project's ability to pull new containers.

It seems like docker pull cannot talk to Docker's servers anymore. Here is the exact error message I am getting:

fatal: [matrix.<domain.tld>]: FAILED! => changed=false 
  attempts: 10
  msg: 'Error pulling image docker.io/devture/exim-relay:4.96-r1-0 - 500 Server Error for http+docker://localhost/v1.43/images/create?tag=4.96-r1-0&fromImage=docker.io%2Fdevture%2Fexim-relay: Internal Server Error ("Get "https://registry-1.docker.io/v2/devture/exim-relay/manifests/sha256:6aeb52edd282561aeaaa41de5757ba714419b21eb3e1454d9658d2bd0313fe14": dial tcp [2600:1f18:2148:bc02:4bf0:3a98:55ae:e3d5]:443: connect: network is unreachable")'

This thread covers it: https://forums.docker.com/t/unable-to-pull-docker-image-due-to-ipv6-since-2023-08-23-around-6pm-cest/137378/8

To Reproduce
Try to pull any container on a machine where Docker has been installed by this Ansible deploy. This can be done by a change in vars.yml or by just running sudo docker pull <container>.

Expected behavior
The container should be pulled.

Matrix Server:

• OS: Debian 12
• Architecture: amd64

Additional context
I am on a brand new install of Debian and attempting to migrate my server to it. I can't run the set up command because it can't pull all the containers it needs. I do not fully understand how this Ansible deploy works, but I did have it install Docker on this server because it doesn't seem to work if Docker was installed by anything else. I hope the information in that Docker thread will help.

[spantaleev]

This only seems to be affecting you.

Try to search for "Linux disable IPv6" and follow any docs you find until you've killed IPv6 off completely.

[MidnightSlicer]

This only seems to be affecting you.

Try to search for "Linux disable IPv6" and follow any docs you find until you've killed IPv6 off completely.

I did that and rebooted the server for good measure. That just changes the error to socket: address family not supported by protocol.

I am honestly at a loss. After undoing the changes it now says connect: network is unreachable again. The only source for yesterday's error code (connect: network is unreachable) was this thread. The only other thing I have found that may be relevant is this issue tracker: https://www.dockerstatus.com/pages/incident/533c6539221ae15e3f000031/64e75cde1670c8052e6ffd6d

How could I apply the changes suggested in the Docker thread (like export DOCKER_BUILDKIT=0) in this environment? Does everything work like a normal Docker install on my server? (I had Ansible install Docker by just running it on a clean server).

[MidnightSlicer]

I can't say I found the issue, but I found another way. It must be some driver thing with my server. I was using Debian 12, but I switched to Ubuntu server (using the same network configs) and it worked immediately.