Files
kubernetes.core/docs/ansible_turbo_mode.rst
T
8c0de50fb8 **This is a manual backport of PR #1242 as merged into main ([031bc47](https://github.com/ansible-collections/kubernetes.core/commit/031bc475fadd0fcf8152d3c9083e22a77e3d870d)).** (#1251)
Emit an explicit deprecation warning when users enable Ansible Turbo mode via `ENABLE_TURBO_MODE`, and stop turbo mode from being used on ansible-core versions where `cloud.common` is not supported.

Turbo mode depends on `cloud.common`, which is being retired. PR #1120 already deprecated the `cloud.common` dependency; without this change, removing that dependency later would cause `ENABLE_TURBO_MODE` to stop working silently. This PR warns users at runtime so they can migrate before support is removed in `kubernetes.core` 8.0.0.

Changes:

* Modules: emit the deprecation at import time via the module-level `deprecate()` in `plugins/module_utils/ansiblemodule.py`. `AnsibleTurboModule.__init__()` ends in `run_on_daemon()`, which calls `exit_json()` and exits the process, so a `deprecate()` call tied to instance construction is never reached.
* Modules: skip turbo mode on ansible-core >= 2.19.0 and fall back to the standard `AnsibleModule`, emitting a warning. `cloud.common` declares `requires_ansible '>=2.15.0,<2.19'`, but that is only enforced at collection load time and the turbo import itself still succeeds on newer cores, so turbo mode would otherwise fail later with `byte indices must be integers or slices, not str`.
* Modules and lookup: warn instead of silently falling back when `ENABLE_TURBO_MODE` is set but `cloud.common` is not installed.
* Lookup: call `Display.deprecated()` in the `k8s` lookup when turbo mode is enabled. The lookup path still works on >= 2.19, so it keeps using turbo and is not version-gated.
* Docs: add deprecation notices to `README.md` and `docs/ansible_turbo_mode.rst`.
* Changelog: add `deprecated_features`, `bugfixes`, and `minor_changes` fragment entries.

On ansible-core < 2.19 with `cloud.common` installed, turbo mode continues to work exactly as before and only gains a deprecation warning. On ansible-core >= 2.19 turbo mode is now deliberately bypassed, which replaces a confusing traceback with a clear warning.

* Bugfix Pull Request
* Docs Pull Request

`ENABLE_TURBO_MODE` / Ansible Turbo mode

On a supported ansible-core (< 2.19) with `cloud.common` installed, turbo mode stays active and users see:

```
[DEPRECATION WARNING]: ENABLE_TURBO_MODE is deprecated, as it relies on the cloud.common collection which is being retired. Setting this environment variable will have no effect once support is removed. This feature will be removed from kubernetes.core in version 8.0.0.
```

On ansible-core >= 2.19.0, turbo mode is bypassed and an additional warning is emitted:

```
[WARNING]: ENABLE_TURBO_MODE is ignored on ansible-core 2.21.4: Ansible Turbo mode requires the cloud.common collection, which supports ansible-core < 2.19.0 only. Continuing without Turbo mode.
```

Warnings are emitted from:

* `plugins/module_utils/ansiblemodule.py` for collection modules
* `plugins/lookup/k8s.py` for the `k8s` lookup plugin

To verify locally:

```shell
export ENABLE_TURBO_MODE=1
ansible localhost -m kubernetes.core.k8s_info -a "kind=Namespace" 2>&1 | grep -iE "deprecat|warning"
```

Integration tests that run with `ENABLE_TURBO_MODE=true` use ansible-core 2.18, where turbo mode remains enabled, so turbo behaviour there is unchanged.

Thanks to @yurnov for diagnosing the unreachable deprecation and for the suggested fix in [mandar242#1](https://github.com/mandar242/kubernetes.core/pull/1).

(cherry picked from commit 031bc475fa)

Co-authored-by: Mandar Kulkarni <mandkulk@redhat.com>
2026-09-15 15:31:27 +01:00

162 lines
5.8 KiB
ReStructuredText

.. _ansible_turbo_mode:
******************
Ansible Turbo mode
******************
Following document provides overview of Ansible Turbo mode in ``kubernetes.core`` collection.
.. contents::
:local:
:depth: 1
.. warning::
Ansible Turbo mode is **deprecated** and will be removed in a future major release
(planned for 8.0.0), as it depends on the ``cloud.common`` collection, which is being
retired. Setting ``ENABLE_TURBO_MODE`` now emits a deprecation warning; once support is
removed, the variable will have no effect. Start planning your migration away from
Turbo mode.
.. warning::
Ansible Turbo mode is supported only with ``ansible-core`` versions **lower than 2.19**.
Starting from ``ansible-core`` 2.19, the ``cloud.common`` collection is no longer supported,
and therefore Ansible Turbo mode is also not supported in ``kubernetes.core``.
If ``ENABLE_TURBO_MODE`` is set to ``1`` (or any truthy value) with ``ansible-core >= 2.19.0``,
Turbo mode is ignored and a warning is emitted; modules run without it.
Synopsis
--------
- A brief introduction about Ansible Turbo mode in ``kuberentes.core`` collection.
- Ansible Turbo mode is an optional performance optimization. It can be enabled by installing the cloud.common collection and setting the ``ENABLE_TURBO_MODE`` environment variable.
Requirements
------------
The following requirement is needed on the host that executes this module.
- The ``cloud.common`` collection (https://github.com/ansible-collections/cloud.common)
You will also need to set the environment variable ``ENABLE_TURBO_MODE=1`` on the managed host. This can be done in the same ways you would usually do so, for example::
---
- hosts: remote
environment:
ENABLE_TURBO_MODE: 1
tasks:
...
Installation
------------
You can install ``cloud.common`` collection using following command::
# ansible-galaxy collection install cloud.common
Current situation without Ansible Turbo mode
============================================
The traditional execution flow of an Ansible module includes the following steps:
- Upload of a ZIP archive with the module and its dependencies
- Execution of the module
- Ansible collects the results once the script is finished
These steps happen for each task of a playbook, and on every host.
Most of the time, the execution of a module is fast enough for
the user. However, sometime the module requires significant amount of time,
just to initialize itself. This is a common situation with the API based modules.
A classic initialization involves the following steps:
- Load a Python library to access the remote resource (via SDK)
- Open a client
- Load a bunch of Python modules.
- Request a new TCP connection.
- Create a session.
- Authenticate the client.
All these steps are time consuming and the same operations will be running again and again.
For instance, here:
- ``import openstack``: takes 0.569s
- ``client = openstack.connect()``: takes 0.065s
- ``client.authorize()``: takes 1.360s,
These numbers are from test running against VexxHost public cloud.
In this case, it's a 2s-ish overhead per task. If the playbook
comes with 10 tasks, the execution time cannot go below 20s.
How Ansible Turbo Module improve the situation
==============================================
``AnsibleTurboModule`` is actually a class that inherites from
the standard ``AnsibleModule`` class that your modules probably
already use.
The big difference is that when a module starts, it also spawns
a little Python daemon. If a daemon already exists, it will just
reuse it.
All the module logic is run inside this Python daemon. This means:
- Python modules are actually loaded one time
- Ansible module can reuse an existing authenticated session.
The background service
======================
The daemon kills itself after 15s, and communication are done
through an Unix socket.
It runs in one single process and uses ``asyncio`` internally.
Consequently you can use the ``async`` keyword in your Ansible module.
This will be handy if you interact with a lot of remote systems
at the same time.
Security impact
===============
``ansible_module.turbo`` open an Unix socket to interact with the background service.
We use this service to open the connection toward the different target systems.
This is similar to what SSH does with the sockets.
Keep in mind that:
- All the modules can access the same cache. Soon an isolation will be done at the collection level (https://github.com/ansible-collections/cloud.common/pull/17)
- A task can load a different version of a library and impact the next tasks.
- If the same user runs two ``ansible-playbook`` at the same time, they will have access to the same cache.
When a module stores a session in a cache, it's a good idea to use a hash of the authentication information to identify the session.
Error management
================
``ansible_module.turbo`` uses exceptions to communicate a result back to the module.
- ``EmbeddedModuleFailure`` is raised when ``json_fail()`` is called.
- ``EmbeddedModuleSuccess`` is raised in case of success and returns the result to the origin module process.
These exceptions are defined in ``ansible_collections.cloud.common.plugins.module_utils.turbo.exceptions``.
You can raise ``EmbeddedModuleFailure`` exception yourself, for instance from a module in ``module_utils``.
.. note:: Be careful with the ``except Exception:`` blocks.
Not only they are bad practice, but also may interface with this
mechanism.
Troubleshooting
===============
You may want to manually start the server. This can be done with the following command:
.. code-block:: shell
PYTHONPATH=$HOME/.ansible/collections python -m ansible_collections.cloud.common.plugins.module_utils.turbo.server --socket-path $HOME/.ansible/tmp/turbo_mode.kubernetes.core.socket
You can use the ``--help`` argument to get a list of the optional parameters.