talos: [1.10.0] etcd won't start after upgrade due to SELinux disabled

Bug Report

Description

Metal platform, old Fujitsu Primergy TX120 S3P server.

After upgrading from 1.9.5 to 1.10.0 (which was slightly botched due to TPM issue #10847) - I disabled the TPM to finish upgrading, and the node started, but etcd won’t start, saying that the SElinux label is specified, but it’s disabled or not supported.

Logs

172.21.100.8: user: warning: [2025-05-03T09:29:43.026356962Z]: [talos] service[etcd](Waiting): Error running Containerd(etcd), going to restart forever: failed to create task: "etcd": failed to create shim task: OCI runtime create failed: runc create failed: selinux label is specified in config, but selinux is disabled or not supported
172.21.100.8: user: warning: [2025-05-03T09:29:48.057332962Z]: [talos] service[etcd](Waiting): Error running Containerd(etcd), going to restart forever: failed to create task: "etcd": failed to create shim task: OCI runtime create failed: runc create failed: selinux label is specified in config, but selinux is disabled or not supported

Environment

  • Talos version: v1.10.0
  • Kubernetes version: v1.31.4
  • Platform: metal

Image

About this issue

  • Original URL
  • State: closed
  • Created 2 years ago
  • Comments: 22 (22 by maintainers)

Most upvoted comments

It seems you’ve a pod that disables selinux from userspace

ugh -_- yes, I think I do, although I can’t remember why the hell I added it. thank you!

Did removing it help? Yes, runtime disable is not supported (in enforcing mode SELinux won’t let anyone do this, but in permissive you still should use selinux=0 to disable it at boot)

Yes, after removing the daemonset and rebooting the node - etcd came on ok. I vaguely remember that talos 1.9 had some issues with cilium CNI or something, that’s when I added the daemonset as a ‘workaround’.

I guess the issue can be closed now?

Thank you! I believe this should be documented, because some systems work with runtime disable (not Red Hat iirc), and users might want to do this. Will document how SELinux should be disabled properly and then close

selinuxState: enabled, permissive

Okay, so SELinux is enabled, however containerd fails as if SELinux were disabled but still requested. Rather weird.

Could you please post a support bundle here (it should not contain any sensitive data, unless you consider local node IPs like 172.21.100.8 to be confidential)