When using a disconnected env and OPENSHIFT_INSTALL_RELEASE_IMAGE_MIRROR env var is specified, the create-cluster-and-infraenv service fails[*].
Seems that the issue happens due to a missing registries.conf in the assisted-service container, which is required for pulling the image.

[*[
create-cluster-and-infraenv[2784]: level=fatal msg="Failed to register cluster with assisted-service: command 'oc adm release info -o template --template '{{.metadata.version}}' --insecure=true quay.io/openshift-release-dev/ocp-release@sha256:3c050cb52fdd3e65c518d4999d238ec026ef724503f275377fee6bf0d33093ab --registry-config=/tmp/registry-config1560177852' exited with non-zero exit code 1: \nerror: unable to read image quay.io/openshift-release-dev/ocp-release@sha256:3c050cb52fdd3e65c518d4999d238ec026ef724503f275377fee6bf0d33093ab: Get "http://quay.io/v2/\": net/http: request canceled while waiting for connection (Client.Timeout exceeded while awaiting headers)\n"

4.14

100%

1. Add registries.conf with mirror config set to a local registry (e.g. use imageContentSources in install-config)
2. Ensure that a custom release image mirror that refers the registry is set on OPENSHIFT_INSTALL_RELEASE_IMAGE_MIRROR env var.
3. Boot the machine on a disconnected env.

create-cluster-and-infraenv service fails pull the release image.

create-cluster-and-infraenv service should finish successfully.

Pushed a PR to the installer for propagating registries.conf: https://github.com/openshift/installer/pull/7332

We have a workaround in the appliance by overriding the service:
https://github.com/openshift/appliance/pull/94/