ansible-runner-service

This project wraps the ansible_runner interface inside a REST API enabling ansible playbooks to be executed and queried from other platforms.

The incentive for this is two-fold;

• provide Ansible integration to non-python projects
• provide a means of programmatically running playbooks where the ansible engine is running on a separate host or in a separate container

Features

The core of this project is ansible_runner, so first of all, a quick call out to those folks for such an awesome tool!

Security

• https support (http not supported)
  • uses self-signed if existing crt/key files are not present
  • if not present, generates self-signed on first start up
• API endpoints
  • /api and /metrics endpoints are deemed public
  • all other endpoints are require a token (JWT) for access
• token based access requires a valid login using http basicauth, to initiate token generation
  • token has a default TTL (24 hours)
• creates or reuses ssh pub/priv keys for communication with target hosts
• supports ip whitelist to further lock down access to the API

Monitoring

• /metrics endpoint provides key metrics for monitoring the instance with prometheus
• a sample Grafana dashboard is provided in the misc/dashboards directory to track activity

Playbook Execution

• exposes playbooks by name found within the project folder
• supports Ansible environments that use private libraries (ie. the library directory is stored within the project folder)
• playbooks can be run with tags to change execution behavior
• playbooks can use limit to restrict actions to a specific host
• running playbooks may be cancelled
• supports execution of concurrent playbooks

Playbook State

• playbook state and output can be queried during and after execution
• playbook state shows overall status, with current active task name
• the caller can request all events associated with current or past playbook runs
• events may be filtered for specific output e.g. ?task=RSEULTS to show events with a taskname of RESULTS

Inventory management

• hosts and ansible groups are managed through the API /groups and /hosts endpoints
• Before a host can be added to the inventory, it is checked for dns, and passwordless ssh
• missing public keys on 'candidate' hosts, result in the instance's public key being returned to the caller. The requester can then arrange for this key to be installed on the candidate host.

Developer Friendly

• simple to use REST API allowing playbooks to be run, and results/state queried
• provides a /api endpoint describing each endpoint
• /api content is automatically generated and has no external dependencies
• each description includes an curl command example, together with output

Deployment

• supports docker - Dockerfile and README included
• cross platform support (docker image uses CentOS7 base, build process executes against Ubuntu)
• can be packaged as an rpm or run as a container
• designed to offer core ansible functionality, supplemented by a users set of playbooks/roles/library
• supports configuration options through a specific /etc directory
• configuration options may be overridden at the command line for diagnostics
• all relevant activity is logged

Prerequisites

So far, testing has been mainly against Fedora (28) and the CentOS7 for the docker image. Other distros may work fine (Travis build uses Ubuntu Trusty for example!).

Package Dependencies

• Python 3.6
• pyOpenSSL (python3-pyOpenSSL on Fedora, CentOS pyOpenSSL)
• ansible_runner 1.1.1 or above

(see requirements.txt for a more complete list of the python dependencies)

if in doubt, look in the misc/docker folder and build the container!

Installation

Try before you buy...assuming you have an environment that meets the python3 dependencies, simply unzip the archive and run :)

python3 ansible_runner_service.py

When you run from any directory outside of /usr, the script regards this as 'dev' mode. In this mode, all files and paths are relative to the path that you've unzipped the project into.

For 'prod' mode, a setup.py is provided. Once the package is installed and called from /usr/*/bin, the script will expect config and output files to be found in all the normal 'production' locations (see proposed file layout below)

sudo python3 setup.py install --record installed_files --single-version-externally-managed

Once this is installed, you may start the service with

ansible_runner_service

API Endpoints

Once the service is running, you can point your browser at https://localhost:5001/api to show which endpoints are available. Each endpoint is described along with a curl example showing invocation and output.

You may click on any row to expand the description of the API route and show the curl example. The app uses a self-signed certificate, so all examples use the -k parameter (insecure mode).

Note: It is not the intent of this API to validate the parameters passed to it. It is assumed that parameter selection and validation happen prior to the API call.

Here's a quick 'cheat sheet' of the API endpoints.

API Route	Description
/api	Show available API endpoints (this page)
/api/v1/groups	List all the defined groups in the inventory
/api/v1/groups/<group_name>	Manage groups within the inventory
/api/v1/hosts	Return a list of hosts from the inventory
/api/v1/hosts/<host_name>	Show group membership for a given host
/api/v1/hosts/<host_name>/groups/<group_name>	Manage ansible control of a given host
/api/v1/jobs/<play_uuid>/events	Return a list of events within a given playbook run (job)
/api/v1/jobs/<play_uuid>/events/<event_uuid>	Return the output of a specific task within a playbook
/api/v1/login	Authenticate user and provide token
/api/v1/playbooks	Return the names of all available playbooks
/api/v1/playbooks/<play_uuid>	Query the state or cancel a playbook run (by uuid)
/api/v1/playbooks/<playbook_name>	Start a playbook by name, returning the play's uuid
/api/v1/playbooks/<playbook_name>/tags/	Start a playbook using tags to control which tasks run
/metrics	Provide prometheus compatible statistics which describe playbook activity

Testing

Testing to date has all been lab based, so please bear this in mind if considering using this tool for production use cases (bug reports welcome!). Playbook integration with Ceph and Gluster has been the primary focus together with the probe-disks.yml playbook. Did you spot the theme?..It's all about the storage™ :)

For example, with ceph the osd-configure.yml playbook has been tested successfully.

Manual Testing

The archive, downloaded from github, contains a simple playbook that just uses the bash sleep command - enabling you to quickly experiment with the API.

Use the steps below (dev mode), to quickly exercise the API

1. Authenticate user and provide token
   curl -k -i --user admin:admin https://localhost:5001/api/v1/login -X get
2. Get the list of available playbooks (should just be test.yml)
   curl -k -i -H "Authorization: eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJleHAiOjE1MzczODA3MTR9.CbTXvBum5mCq9s56wJNiMn8JLJ0UzzRdwdeOFctJtbI" https://localhost:5001/api/v1/playbooks -X GET
3. Run the test.yml playbook, passing the time_delay parameter (30 secs should be enough).
   curl -k -i -H "Authorization: eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJleHAiOjE1MzczODA3MTR9.CbTXvBum5mCq9s56wJNiMn8JLJ0UzzRdwdeOFctJtbI" -H "Content-Type: application/json" --data '{"time_delay": 30}' https://localhost:5001/api/v1/playbooks/test.yml -X POST
4. The previous command will return the playbooks UUID. Use this identifier to query the state or progress of the run.
   curl -k -i -H "Authorization: eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJleHAiOjE1MzczODA3MTR9.CbTXvBum5mCq9s56wJNiMn8JLJ0UzzRdwdeOFctJtbI" https://localhost:5001/api/v1/playbooks/f39069aa-9f3d-11e8-852f-c85b7671906d -X GET
5. Get a list of all the events in a playbook. The return list consists of all the job event ID's
   curl -k -i -H "Authorization: eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJleHAiOjE1MzczODA3MTR9.CbTXvBum5mCq9s56wJNiMn8JLJ0UzzRdwdeOFctJtbI" https://localhost:5001/api/v1/jobs/f39069aa-9f3d-11e8-852f-c85b7671906d/events -X GET
6. To get specific output from a job event, you can query the job event
   curl -k -i -H "Authorization: eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJleHAiOjE1MzczODA3MTR9.CbTXvBum5mCq9s56wJNiMn8JLJ0UzzRdwdeOFctJtbI" https://localhost:5001/api/v1/jobs/f39069aa-9f3d-11e8-852f-c85b7671906d/events/13-c85b7671-906d-e52d-d421-000000000008 -X GET

Obviously you'll need to change the token, play and job uuids for your run :)

Tips & Tricks

1. Tweaking the environment:The script uses a configuration module which is accessible across the different modules within the project. There are two ways that settings in the configuration module can be overridden;

   • by using a config.yaml file
   • by providing a setting value when starting the ansible_runner_service program

2. Overriding configuration at run time, lets you do quick tests like this;

   • start the service, but don't perform any passwordless ssh tests

   $ ssh_checks=false python3 ansible_runner_service
   

   • change the target user when validating ssh connection is in place

   $ target_user=root python3 ansible_runner_service
   

Automated Build & Testing

The project uses Travis CI integration to check the following;

• Installation
• code style (using flake8)
• Ansible inventory management (groups/hosts)
• API endpoints using test data and a test playbook

For more info, look at the .travis.yml file.

File Layout (Proposed)

/etc/ansible-runner-service

• logging.yaml
• config.yaml
• ansible-runner-service.crt
• ansible-runner-service.key

/usr/share/ansible-runner-service

• artifacts
• inventory
• env
• project
  • roles (optional)
  • library (optional)
  • test.yaml
• roles

/var/log/ansible-runner-service.log

/usr/share/doc/ansible-runner-service

• README.md
• LICENSE.md

/etc/systemd/system

• ansible-runner-service.service

/usr/bin/ or /usr/local/bin

• ansible_runner_service