Docker Proxy Changes Applied to the Wrong Daemon

You added a proxy to docker.service, restarted it, and the next pull behaved exactly as before. On a machine with both apt and snap installations, check whether that is the daemon your CLI is using.

There are three separate identities to match: the CLI executable, the selected endpoint, and the server process behind that endpoint. Finding one docker command in PATH settles only the first.

docker context show
docker context inspect

Docker contexts describe endpoints. DOCKER_HOST, DOCKER_CONTEXT, and command-line endpoint selections can override the apparent default, so inspect the environment and invocation too.

Then associate the selected socket or remote address with the actual daemon and its service configuration. Two installations can have different units, configuration paths, and storage. A remote context may point off the machine entirely.

Apply the proxy through the interface belonging to that daemon. Docker's documented daemon settings and its startup environment are not a broadcast to every Docker installation on the host.

Do this identification before uninstalling a package to “clean things up.” An unused-looking daemon may own containers or images that the other instance does not manage. The immediate proxy repair is to configure the selected server deliberately; deciding which installation to retain is a separate maintenance change.