Summary
On CI Windows WSL, starting containers fails with a missing cgroup pids controller. Hyper-V with the same Podman version passes. Local rootful WSL also passes.
This is not Kind-specific — Kind create is one manifestation; basic container start fails the same way.
Confirmed
k8s E2E (Kind)
https://github.com/podman-desktop/e2e/actions/runs/36719001023 (Podman v6.1.3)
- Win11 WSL: Kind create fails
- Win10/11 Hyper-V: 57/57 pass
crun: controller `pids` is not available under
/sys/fs/cgroup/non-systemd/machine.slice/libpod-....scope/container/cgroup.controllers
OCI runtime error (exit 126)
Kind pulls kindest/node, then podman run for the control-plane fails. Playwright timeouts waiting for "Go back to resources" are a consequence.
Main WSL E2E (smoke — latest + nightly)
https://github.com/podman-desktop/e2e/actions/runs/36696478637
- All 4 jobs fail: Win10/11 x latest (v6.1.3) / nightly (main-c2f2b19)
- Same error on basic alpine container start, pods, network, etc.
- Seen under both
machine.slice (rootful) and user.slice (rootless)
Example (container-smoke):
Starting the container threw an error: (HTTP code 500)
crun: controller `pids` is not available under
/sys/fs/cgroup/non-systemd/machine.slice/.../cgroup.controllers
Local contrast
Rootful WSL: pids present in cgroup.controllers; Kind create and alpine container start succeed. Issue is CI WSL-specific, not a Podman Desktop / Playwright bug.
Not confirmed
Why nested cgroups on MAPT WSL lack pids (cgroup manager, controller delegation, WSL/kernel config, machine defaults).
Next checks on a failing WSL runner
podman info → cgroupManager, cgroupVersion, cgroupControllers
cat /sys/fs/cgroup/cgroup.controllers
- Same on a green Hyper-V job for comparison
Summary
On CI Windows WSL, starting containers fails with a missing cgroup
pidscontroller. Hyper-V with the same Podman version passes. Local rootful WSL also passes.This is not Kind-specific — Kind create is one manifestation; basic container start fails the same way.
Confirmed
k8s E2E (Kind)
https://github.com/podman-desktop/e2e/actions/runs/36719001023 (Podman v6.1.3)
Kind pulls
kindest/node, thenpodman runfor the control-plane fails. Playwright timeouts waiting for "Go back to resources" are a consequence.Main WSL E2E (smoke — latest + nightly)
https://github.com/podman-desktop/e2e/actions/runs/36696478637
machine.slice(rootful) anduser.slice(rootless)Example (container-smoke):
Local contrast
Rootful WSL:
pidspresent incgroup.controllers; Kind create and alpine container start succeed. Issue is CI WSL-specific, not a Podman Desktop / Playwright bug.Not confirmed
Why nested cgroups on MAPT WSL lack
pids(cgroup manager, controller delegation, WSL/kernel config, machine defaults).Next checks on a failing WSL runner
podman info→ cgroupManager, cgroupVersion, cgroupControllerscat /sys/fs/cgroup/cgroup.controllers