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
{{ message }}
Repository navigation
test(tmachine): prepare runner environments for the e2e/rust migration #5275
I maintain OpenShell's end-to-end coverage and am moving the e2e/rust suite onto testkit and the tmachine runner. Before those tests can move, every tmachine runner environment needs the same basics the old lanes had, so ported tests behave the same on Docker, Podman, k3s and k3s-ha.
Description
Two gaps block the port:
host.openshell.internal on k3s. Docker and Podman provide this alias natively. The tmachine k3s and k3s-ha installs don't set the Helm value server.hostGatewayIP, so sandboxes there can't reach a server the test starts on the VM. The e2e-host-gateway tests and the Kubernetes shared-supervisor tests depend on it.
Python workload image in every environment. Many e2e/rust tests run python3 in sandboxes. tmachine already builds openshell/e2e-python:dev but only loads it for the Podman e2e testsuite. Load it on every installer (Docker, Podman, k3s containerd) and give testkit an env var and helper for it next to OPENSHELL_TEST_BASE_IMAGE. The base image default doesn't change.
A small features test proves both: a sandbox running the Python image reaches a server the test starts on the host through host.openshell.internal, on every runner row including k3s and k3s-ha.
Activity