You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This comment sheds some light as to what we can do here. This should work, I just need to figure out how to bind to the host volume from within the DOOD (docker on outer docker) instance.
codespaces environments are actually containers. We can get access to the containing host-instance of docker through the bind-mounted unix socket at /var/run/docker-host.sock. We can call into it like curl --unix-socket /var/run/docker-host.sock -X GET http://localhost/containers/json and see other containers on the same host.
The issue with the above, is that there is no way for the sibling containers to talk to each other.
So, if we want to continue with the leveraging the docker sock, we need some way to resolve the networking.
Other than utilizing the already bound docker socket, I can see two other alternatives.
The issue with this? This could be a single transfer of a command, or it could be a nightmare to configure. 🤷♂️
Run the test-proxy via the command, and not docker, while running on docker.
Means that the startup script will need a "if run using test-proxy instead of the docker command"
Ideally, we are able to magic the environment to get it to the point where there is simply 0 thought, and we can stand up the instance the same way in codespaces that we do locally. For that to work, the dind approach is the way to go.
Repro BranchCommit at time of issue creationIn Main now!
If spun up within codespaces, we are still forced to
docker cp
out of the instance.I'm very concerned that due to the fact that this is docker running within docker, there won't be a valid way to map the volume. We will see though.
The text was updated successfully, but these errors were encountered: