K3s fatal error

The steps to follow when the k3s service won’t start

Steps / Information Gathering

Check the config

Resolve of fail of a k3s check-config

k3s check-config
# you can ignore modprobe: FATAL: Module configs not found (see above)

Example of fail

Generally Necessary:
- apparmor: enabled, but apparmor_parser missing (fail)
    (use "apt-get install apparmor" to fix this)

Note that

modprobe: FATAL: Module configs not found in directory /lib/modules/6.1.0-23-amd64

where: 6.1.0-23-amd64 is the output of uname -r

It’s not really fatal, this is a red herring. There is no kernel module missing. System verification needs to find the kernel configuration file and canot find it.

If:

• the configuration file cannot be found,
• there is no configs module, > system verification check will just error out.

See: https://github.com/kubernetes/kubernetes/issues/41025

Check the system service unit

Check the validity of the system service unit.

Grab the k3s command issued and execute it on a shell.

cat /etc/systemd/system/k3s.service

Example:

ExecStart=/usr/local/bin/k3s server '--disable=traefik'

Journal / Status / Log

When running under Systemd, logs are sent to Journald and can be viewed using journalctl.

sudo systemctl status k3s.service
# Last 50
journalctl -u k3s.service -n 50
# tail
journalctl -u k3s.service -f
# log since the last machine boot
journalctl -u k3s.service -b

You can generate more detailed logs by using the --debug flag when starting K3s (or debug: true in the configuration file).

Get more information on the play with the debug flag

You can add the --debug flag to see all variables.

kubee --debug --cluster "$KUBEE_CLUSTER_NAME" cluster play

Check the known-issues

If no clue: https://docs.k3s.io/known-issues

Once resolved

Check alive / cluster info

# on server
k3s kubectl cluster-info
# on client
kubee --cluster "$KUBEE_CLUSTER_NAME" kubectl cluster-info