Re: [PATCH v3 2/3] declared_hosts: add support for pre-existing infrastructure

On 03/09/2025 01.53, Luis Chamberlain wrote:
> This adds support for using pre-existing infrastructure (bare metal servers,
> pre-provisioned VMs, cloud instances) that users already have SSH access to,
> bypassing the kdevops bringup process.
> 
> We borrow the DECLARE_* foo practice from the Linux kernel to ensure the
> user will declare the hosts they already have set up with:
> 
>   make DECLARE_HOSTS="foo bar" defconfig-foo
>   or
>   make DECLARE_HOSTS="foo bar" menuconfig
> 
> We just skip the data partition setup, at the role level. The user
> is encouraged to set DATA_PATH if they want something other than /data/
> to be expected to be used. The onus is on them to ensure that the
> DATA_PATH works for the user the the host is configured to ssh access
> to already.
> 
> Currently no workflows are fully supported with declared hosts.
> Each workflow requires individual review and testing to ensure proper
> operation with pre-existing infrastructure before being enabled.
> 
> Generated-by: Claude AI
> Signed-off-by: Luis Chamberlain <[email protected]>
> ---

...

> diff --git a/kconfigs/Kconfig.declared_hosts b/kconfigs/Kconfig.declared_hosts
> new file mode 100644
> index 00000000..cfdae8fa
> --- /dev/null
> +++ b/kconfigs/Kconfig.declared_hosts
> @@ -0,0 +1,71 @@
> +# Configuration for declared hosts that skip bringup process
> +
> +config KDEVOPS_USE_DECLARED_HOSTS
> +	bool "Use declared hosts (skip bringup process)"
> +	select WORKFLOW_INFER_USER_AND_GROUP
> +	output yaml
> +	help
> +	  Enable this option to use pre-existing hosts that you have already
> +	  configured with SSH access. This is useful for:
> +
> +	  * Bare metal systems
> +	  * Pre-provisioned VMs or cloud instances
> +	  * Systems managed by other infrastructure tools
> +
> +	  When this option is enabled:
> +	  - SSH keys will not be generated (assumes you already have access)
> +	  - Bringup and teardown operations will be skipped
> +	  - User and group settings will be inferred from the target hosts
> +	  - You must provide the list of hosts in KDEVOPS_DECLARED_HOSTS
> +
> +	  This option automatically:
> +	  - Selects WORKFLOW_INFER_USER_AND_GROUP to detect the correct
> +	    user and group on the target systems
> +	  - Assumes SSH access is already configured
> +
> +if KDEVOPS_USE_DECLARED_HOSTS
> +
> +config KDEVOPS_DECLARED_HOSTS
> +	string "List of declared hosts"
> +	output yaml
> +	default "$(shell, echo ${DECLARED_HOSTS})"
> +	default ""
> +	help
> +	  Provide a list of hostnames or IP addresses for the pre-existing
> +	  systems you want to use. These hosts must already be accessible
> +	  via SSH with the appropriate keys configured.
> +
> +	  Format: Space or comma-separated list
> +	  Example: "host1 host2 host3" or "host1,host2,host3"
> +
> +	  These hosts will be used directly without any bringup process.
> +	  Make sure you have:
> +	  - SSH access configured
> +	  - Required packages installed
> +	  - Appropriate user permissions
> +
> +config KDEVOPS_DECLARED_HOSTS_PYTHON_INTERPRETER
> +	string "Python interpreter path on declared hosts"
> +	default "/usr/bin/python3"
> +	output yaml
> +	help
> +	  Specify the path to the Python interpreter on the declared hosts.
> +	  This is required for Ansible to function properly.
> +
> +	  Common values:
> +	  - /usr/bin/python3 (most modern systems)
> +	  - /usr/bin/python (older systems)
> +	  - /usr/local/bin/python3 (custom installations)

Isn't this redundant?

In the hosts template files, this patch does this:

ansible_python_interpreter = "{{ kdevops_declared_hosts_python_interpreter | default(kdevops_python_interpreter) }}"

I think, using the kdevops_python_interperter config/yaml directly should be
sufficient. Or what I am missing?