runc sometimes fails with "no such file or directory"

[prezha]

What Happened?

we see these in some kvm+containerd tests:

example 1:

	I0116 02:30:24.180449  339661 ssh_runner.go:195] Run: sudo runc --root /run/containerd/runc/k8s.io list -f json
...
	stderr:
	time="2024-01-16T02:30:24Z" level=error msg="stat /run/containerd/runc/k8s.io/4744173521ad7687144254eea07ff8aff4223eee041b7cec0b7bda79fc00a1d0: no such file or directory"

example 2:

	I0116 01:58:56.995758  567664 ssh_runner.go:195] Run: sudo runc --root /run/containerd/runc/k8s.io list -f json
...
	stderr:
	time="2024-01-16T01:58:57Z" level=error msg="stat /run/containerd/runc/k8s.io/7010e3c493ae1d4af36adeaef2828f0f9da1b15cdf9f78f705d7c137e0446ba3: no such file or directory"

we use runc v1.1.10, and once released, some future version could possibly solve this issue:

(our) runc v1.1.10: https://github.com/opencontainers/runc/blob/18a0cb0f32bcac2ecc9a10f327d282759c144dab/list.go#L131-L134

			st, err := os.Stat(filepath.Join(absRoot, item.Name()))
			if err != nil {
				fatal(err)
			}

(latest) runc v1.1.11 - unchanged: https://github.com/opencontainers/runc/blob/4bccb38cc9cf198d52bebf2b3a90cd14e7af8c06/list.go#L131-L134

			st, err := os.Stat(filepath.Join(absRoot, item.Name()))
			if err != nil {
				fatal(err)
			}

(current) runc HEAD: https://github.com/opencontainers/runc/blob/0c5a73535503216a8c15a86aa9022d3b0d995994/list.go#L129-L136

		st, err := item.Info()
		if err != nil {
			if errors.Is(err, os.ErrNotExist) {
				// Possible race with runc delete.
				continue
			}
			return nil, err
		}

so, it's due to a Possible race with runc delete

interestingly, this change was merged in pr 3379 by @kolyshkin on 2022-01-27, but it looks like it was never released (see the current "runc v1.1.11" above, released on 2024-01-02)

we could simply retry until this gets fixed/released upstream, and then we update to that runc version

[k8s-triage-robot]

The Kubernetes project currently lacks enough contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle stale
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

[kolyshkin]

runc 1.1 backport: opencontainers/runc#4231 (will be released in 1.1.13 if/when we do another 1.1.x release)

[prezha]

/remove-lifecycle stale