Documentation how validations work

Start importing the TripleO Ansible validations from Clapper repository
[1] into tripleo-validations.

This commit lays down the structure for the ansible validations and
documents how to write and run validations.

Next commits will import existing validations one by one.


Change-Id: I946d4e61bda1382c5299c7494fb994d870cdc56b
Co-Authored-By: Tomas Sedovic <>
This commit is contained in:
Martin André 2016-05-27 15:15:50 +02:00
parent 8ed8cc11ab
commit 438aa71711
3 changed files with 402 additions and 0 deletions

View File

@ -22,3 +22,368 @@ Prerequisites
The TripleO validations require Ansible 2.0 or above::
$ sudo pip install 'ansible>=2'
Existing validations
Here are all the validations that currently exist. They're grouped by
the deployment stage they're should be run on.
Validations can belong to multiple groups.
Validations that are run on a fresh machine *before* the undercloud is
Pre Introspection
Validations that are run when the undercloud is ready to perform hardware
Pre Deployment
Validation that are run right before deploying the overcloud.
Post Deployment
Validations that are run after the overcloud deployment finished.
Writing Validations
All validations are written in standard Ansible with a couple of extra
meta-data to provide information to the Mistral validation framework.
For people not familiar with Ansible, get started with their `excellent
documentation <>`_.
After the generic explanation on writing validations is a couple of concrete
Directory Structure
All validations are located in the ``validations`` directory. It
contains a couple of subdirectories:
- the ``files`` directory contains scripts that are directly executable;
- the ``library`` one is for custom Ansible modules available to the
- ``tasks`` is for common steps that can be shared between the validations.
Here is what the tree looks like::
├── first_validation.yaml
├── second_validation.yaml
├── files
│ └──
├── library
│ ├──
│ └──
└── tasks
└── some_task.yaml
Sample Validation
Each validation is an Ansible playbook with a known location and some
meta-data. Here is what a minimal validation would look like::
- hosts: overcloud
name: Hello World
description: This validation prints Hello World!
- name: Run an echo command
command: echo Hello World!
It should be saved as ``validations/hello_world.yaml``.
As shown here, the validation playbook requires three top-level directives:
``hosts``, ``vars -> metadata`` and ``tasks``.
``hosts`` specify which nodes to run the validation on. Based on the
``hosts.sample`` structure, the options can be ``all`` (run on all nodes),
``undercloud``, ``overcloud`` (all overcloud nodes), ``controller`` and
The ``vars`` section serves for storing variables that are going to be
available to the Ansible playbook. The validations API uses the ``metadata``
section to read each validation's name and description. These values are then
reported by the API and shown in the UI.
The validations can be grouped together by specifying a ``groups`` metadata.
Groups function similar to tags and a validation can thus be part of many
groups. Here is, for example, how to have a validation be part of the
`pre-deployment` and `hardware` groups::
- pre-deployment
- hardware
``tasks`` contain a list of Ansible tasks to run. Each task is a YAML
dictionary that must at minimum contain a name and a module to use.
Module can be any module that ships with Ansible or any of the custom
ones in the ``library`` subdirectory.
The `Ansible documentation on playbooks
<>`__ provides more detailed
Ansible Inventory
Hosts file
The static inventory file lets you describe your environment. It should look
something like this::
It will have a ``[group]`` section for each role (``undercloud``,
``controller``, ``compute``) listing all the nodes belonging to that group. It
is also possible to create a group from other groups as done with
``[overcloud:children]`` in the above example. If a validation specifies
``hosts: overcloud``, it will be run on any node that belongs to the
``compute`` or ``controller`` groups. If a node happens to belong to both, the
validation will only be run once.
Lastly, there is an ``[all:vars]`` section where to configure certain
Ansible-specific options.
``ansible_ssh_user`` will specify the user Ansible should SSH as. If that user
does not have root privileges, it is possible to instruct it to use ``sudo`` by
setting ``ansible_sudo`` to ``true``.
Learn more at the `Ansible documentation page for the Inventory
Custom Modules
In case the `available Ansible modules
<>`__ don't cover your
needs, it is possible to write your own. Modules belong to the
``validations/library`` directory.
Here is a sample module that will always fail::
#!/usr/bin/env python
from ansible.module_utils.basic import *
if __name__ == '__main__':
module = AnsibleModule(argument_spec={})
module.fail_json(msg="This module always fails.")
Save it as ``validations/library/`` and use it in a validation like
... # some tasks
- name: Running my custom module
... # some other tasks
The name of the module in the validation ``my_module`` must match the file name
(without extension): ````.
The custom modules can accept parameters and do more complex reporting. Please
refer to the guide on writing modules in the Ansible documentation.
Learn more at the `Ansible documentation page about writing custom modules
Running a validation
Running the validations require ansible and a set of nodes to run them against.
These nodes need to be reachable from the operator's machine and need to have
an account it can ssh to and perform passwordless sudo.
The nodes need to be present in the static inventory file.
In general, Ansible and the validations will be located on the *undercloud*,
because it should have connectivity to all the *overcloud* nodes is already set
up to SSH to them.
$ source ~/stackrc
$ ansible-playbook -i hosts path/to/validation.yaml
Example: Verify Undercloud RAM requirements
The Undercloud has a requirement of 16GB RAM. Let's write a validation
that verifies this is indeed the case before deploying anything.
Let's create ``validations/undercloud-ram.yaml`` and put some metadata
in there::
- hosts: undercloud
name: Minimum RAM required on the undercloud
description: >
Make sure the undercloud has enough RAM.
- prep
- pre-introspection
The ``hosts`` key will tell which server should the validation run on. The
common values are ``undercloud``, ``overcloud`` (i.e. all overcloud nodes),
``controller`` and ``compute`` (i.e. just the controller or the compute nodes).
The ``name`` and ``description`` metadata will show up in the API and the
TripleO UI so make sure to put something meaningful there. The ``groups``
metadata applies a tag to the validation and allows to group them together in
order to perform group operations, such are running them all in one call.
Now let's add an Ansible task to test that it's all set up properly. Add
this under the same indentation as ``hosts`` and ``vars``::
- name: Test Output
debug: msg="Hello World!"
When running it, it should output something like this::
$ ansible-playbook -i hosts validations/undercloud-ram.yaml
PLAY [undercloud] *************************************************************
GATHERING FACTS ***************************************************************
ok: [localhost]
TASK: [Test Output] ***********************************************************
ok: [localhost] => {
"msg": "Hello World!"
PLAY RECAP ********************************************************************
localhost : ok=2 changed=0 unreachable=0 failed=0
Writing the full validation code is quite easy in this case because Ansible has
done all the hard work for us already. We can use the ``ansible_memtotal_mb``
fact to get the amount of RAM (in megabytes) the tested server currently has.
For other useful values, run ``ansible -i hosts undercloud -m setup``.
So, let's replace the hello world task with a real one::
- name: Verify the RAM requirements
fail: msg="The RAM on the undercloud node is {{ ansible_memtotal_mb }} MB, the minimal recommended value is 16 GB."
failed_when: "({{ ansible_memtotal_mb }}) < 16000"
Running this, we see::
TASK: [Verify the RAM requirements] *******************************************
failed: [localhost] => {"failed": true, "failed_when_result": true}
msg: The RAM on the undercloud node is 8778 MB, the minimal recommended value is 16 GB.
Because our Undercloud node really does not have enough RAM. Your mileage may
Either way, the validation works and reports the lack of RAM properly!
``failed_when`` is the real hero here: it evaluates an Ansible expression (e.g.
does the node have more than 16 GB of RAM) and fails when it's evaluated as
The ``fail`` line right above it lets us print a custom error in case of
a failure. If the task succeeds (because we do have enough RAM), nothing will
be printed out.
Now, we're almost done, but there are a few things we can do to make this nicer
on everybody.
First, let's hoist the minimum RAM requirement into a variable. That way we'll
have one place where to change it if we need to and we'll be able to test the
validation better as well!
So, let's call the variable ``minimum_ram_gb`` and set it to ``16``. Do this in
the ``vars`` section::
name: ...
description: ...
groups: ...
minimum_ram_gb: 16
Make sure it's on the same indentation level as ``metadata``.
Then, update ``failed_when`` like this::
failed_when: "({{ ansible_memtotal_mb }}) < {{ minimum_ram_gb|int * 1024 }}"
And ``fail`` like so::
fail: msg="The RAM on the undercloud node is {{ ansible_memtotal_mb }} MB, the minimal recommended value is {{ minimum_ram_gb|int * 1024 }} MB."
And re-run it again to be sure it's still working.
One benefit of using a variable instead of a hardcoded value is that we can now
change the value without editing the yaml file!
Let's do that to test both success and failure cases.
This should succeed but saying the RAM requirement is 1 GB::
ansible-playbook -i hosts validations/undercloud-ram.yaml -e minimum_ram_gb=1
And this should fail by requiring much more RAM than is necessary::
ansible-playbook -i hosts validations/undercloud-ram.yaml -e minimum_ram_gb=128
(the actual values may be different in your configuration -- just make sure one
is low enough and the other too high)
And that's it! The validation is now finished and you can start using it in
For reference, here's the full validation::
- hosts: undercloud
name: Minimum RAM required on the undercloud
description: Make sure the undercloud has enough RAM.
- prep
- pre-introspection
minimum_ram_gb: 16
- name: Verify the RAM requirements
fail: msg="The RAM on the undercloud node is {{ ansible_memtotal_mb }} MB, the minimal recommended value is {{ minimum_ram_gb|int * 1024 }} MB."
failed_when: "({{ ansible_memtotal_mb }}) < {{ minimum_ram_gb|int * 1024 }}"

hosts.sample Normal file
View File

@ -0,0 +1,33 @@
# Your undercloud servers. This can be `localhost` if you cloned the
# repository on the undercloud:
#put username here if different than compute nodes, ie stack user
#put username here if different than compute nodes, ie stack user
# Username to SSH as:
# Set to `true` if the SSH user is not root. Ansible will run `sudo`
# for commands requiring root:

View File

@ -22,6 +22,10 @@ classifier =
packages =
data_files =
share/openstack-tripleo-validations/ = hosts.sample
share/openstack-tripleo-validations/validations = validations/*
source-dir = doc/source
build-dir = doc/build