feature: Validate config before restarting rke2

[mueller-ma]

Summary

I had the following invalid ansible-role-rke2 config which lead to a broken /etc/rancher/rke2/config.yaml on the server:

rke2_server_options:
  service-node-port-range: 20000-32767

Maybe the config.yaml can be validated before trying to restart any service?

Issue Type

Feature Idea

[mueller-ma]

@MonolithProjects Would you accept an PR that installs yamllint on the node and uses that to validate /etc/rancher/rke2/config.yaml? It can be made optional to avoid the installation of yamllint.

[MonolithProjects]

Hi, yamllint will not catch this. As far as i know, currently it is not possible to check RKE2 config prior the service start/restart.

[mueller-ma]

In this case yamllint will catch the issue:

$ yamllint /etc/rancher/rke2/config.yaml -d relaxed
/etc/rancher/rke2/config.yaml
  1:1       warning  too many blank lines (1 > 0)  (empty-lines)
  16:1      error    syntax error: could not find expected ':' (syntax)

(Warning can be ignored)

[MonolithProjects]

But you wrote that your problem was caused by

rke2_server_options:
  service-node-port-range: 20000-32767

I thought you are talking about some kind of port conflict because of you changed the ports range. And of course this can not cause yaml syntax issue.

What was the problem in the line no.16 ?

[mueller-ma]

The following yaml in Ansible causes an invalid yaml file on the server:

# On Ansible controller
rke2_server_options:
  service-node-port-range: 20000-32767

# On rke2 server
$ sudo grep service-node-port-range /etc/rancher/rke2/config.yaml
service-node-port-range:
# Invalid yaml here. Causes 'syntax error: could not find expected ':' (syntax)'

Correct is:

# On Ansible controller
rke2_server_options:
  - "service-node-port-range: '20000-32767'"

# On rke2 server
$ sudo grep service-node-port-range /etc/rancher/rke2/config.yaml
service-node-port-range: '20000-32767' # Valid yaml here

[MonolithProjects]

Ah OK got it now. rke2_server_options variable has to be a list. To make sure we are using correct variable type we should use assert module and introduce task which will check them. I am reopening this issue.

[mueller-ma]

Indeed, that's better than using yamllint on the final config. Maybe use Role argument validation instead of adding a task.