GeezerVison

Workflow Service for OpenStack cloud. This project aims to provide a mechanism to define tasks and workflows without writing code, manage and execute them in the cloud environment.

The following are the steps to install Mistral on debian-based systems.

To install Mistral, you have to install the following prerequisites:

$ apt-get install python-dev python-setuptools libffi-dev \
  libxslt1-dev libxml2-dev libyaml-dev libssl-dev

Mistral can be used without authentication at all or it can work with OpenStack.

In case of OpenStack, it works only with Keystone v3, make sure Keystone v3 is installed.

First of all, clone the repo and go to the repo directory:

$ git clone https://git.openstack.org/openstack/mistral.git
$ cd mistral

Devstack installation

Information about how to install Mistral with devstack can be found here.

Mistral configuration is needed for getting it work correctly with and without an OpenStack environment.

1. Install and configure a database which can be MySQL or PostgreSQL (SQLite can't be used in production.). Here are the steps to connect Mistral to a MySQL database.

   • Make sure you have installed mysql-server package on your Mistral machine.

   • Install MySQL driver for python:

     $ pip install mysql-python
     

     or, if you work in virtualenv, run:

     $ tox -evenv -- pip install mysql-python
     

     NOTE: If you're using Python 3 then you need to install mysqlclient instead of mysql-python.

   • Create the database and grant privileges:

     $ mysql -u root -p
     mysql> CREATE DATABASE mistral;
     mysql> USE mistral
     mysql> GRANT ALL PRIVILEGES ON mistral.* TO 'mistral'@'localhost' \
            IDENTIFIED BY 'MISTRAL_DBPASS';
     mysql> GRANT ALL PRIVILEGES ON mistral.* TO 'mistral'@'%' IDENTIFIED BY 'MISTRAL_DBPASS';
     

2. Generate mistral.conf file:

   $ oslo-config-generator --config-file tools/config/config-generator.mistral.conf \
     --output-file etc/mistral.conf.sample
   

3. Copy service configuration files:

   $ sudo mkdir /etc/mistral
   $ sudo chown `whoami` /etc/mistral
   $ cp etc/event_definitions.yml.sample /etc/mistral/event_definitions.yml
   $ cp etc/logging.conf.sample /etc/mistral/logging.conf
   $ cp etc/policy.json /etc/mistral/policy.json
   $ cp etc/wf_trace_logging.conf.sample /etc/mistral/wf_trace_logging.conf
   $ cp etc/mistral.conf.sample /etc/mistral/mistral.conf
   

4. Edit file /etc/mistral/mistral.conf according to your setup. Pay attention to the following sections and options:

   [oslo_messaging_rabbit]
   rabbit_host = <RABBIT_HOST>
   rabbit_userid = <RABBIT_USERID>
   rabbit_password = <RABBIT_PASSWORD>
   
   [database]
   # Use the following line if *PostgreSQL* is used
   # connection = postgresql://<DB_USER>:<DB_PASSWORD>@localhost:5432/mistral
   connection = mysql://<DB_USER>:<DB_PASSWORD>@localhost:3306/mistral
   

5. If you are not using OpenStack, add the following entry to the /etc/mistral/mistral.conf file and skip the following steps:

   [pecan]
   auth_enable = False
   

6. Provide valid keystone auth properties:

   [keystone_authtoken]
   auth_uri = http://keystone-host:port/v3
   auth_url = http://keystone-host:port
   auth_type = password
   username = <user>
   password = <password>
   user_domain_name = <usually 'Default'>
   project_name = <project name>
   project_domain_name = <usually 'Default'>
   

7. Register Mistral service and Mistral endpoints on Keystone:

   $ MISTRAL_URL="http://[host]:[port]/v2"
   $ openstack service create --name mistral workflowv2
   $ openstack endpoint create mistral public $MISTRAL_URL
   $ openstack endpoint create mistral internal $MISTRAL_URL
   $ openstack endpoint create mistral admin $MISTRAL_URL
   

8. Update the mistral/actions/openstack/mapping.json file which contains all available OpenStack actions, according to the specific client versions of OpenStack projects in your deployment. Please find more detailed information in the tools/get_action_list.py script.

After local installation you will find the commands mistral-server and mistral-db-manage available in your environment. The mistral-db-manage command can be used for migrating database schema versions. If Mistral is not installed in system then this script can be found at mistral/db/sqlalchemy/migration/cli.py, it can be executed using Python command line.

To update the database schema to the latest revision, type:

$ mistral-db-manage --config-file <path_to_config> upgrade head

To populate the database with standard actions and workflows, type::

$ mistral-db-manage --config-file <path_to_config> populate

For more detailed information about mistral-db-manage script please check file mistral/db/sqlalchemy/migration/alembic_migrations/README.md.

To run Mistral API server:

$ tox -evenv -- python mistral/cmd/launch.py --server api --config-file <path_to_config>

To run Mistral Engine:

$ tox -evenv -- python mistral/cmd/launch.py --server engine --config-file <path_to_config>

To run Mistral Task Executor instance:

$ tox -evenv -- python mistral/cmd/launch.py --server executor --config-file <path_to_config>

Note that at least one Engine instance and one Executor instance should be running in order for workflow tasks to be processed by Mistral.

If you want to run some tasks on specific executor, the task affinity feature can be used to send these tasks directly to a specific executor. You can edit the following property in your mistral configuration file for this purpose:

[executor]
host = my_favorite_executor

After changing this option, you will need to start (restart) the executor. Use the target property of a task to specify the executor:

... Workflow YAML ...
task1:
  ...
  target: my_favorite_executor
... Workflow YAML ...

To run more than one server (API, Engine, or Task Executor) on the same process:

$ tox -evenv -- python mistral/cmd/launch.py --server api,engine --config-file <path_to_config>

The value for the --server option can be a comma-delimited list. The valid options are all (which is the default if not specified) or any combination of api, engine, and executor.

It's important to note that the fake transport for the rpc_backend defined in the configuration file should only be used if all Mistral servers are launched on the same process. Otherwise, messages do not get delivered because the fake transport is using an in-process queue.

1. Complete Mistral documentation.

   Mistral documentation should be more usable. It requires focused work to make it well structured, eliminate gaps in API/Mistral Workflow Language specifications, add more examples and tutorials.

   Definition of done: All capabilities are covered, all documentation topics are written using the same style and structure principles. The obvious sub-goal of this goal is to establish these principles.

2. Finish Mistral multi-node mode.

   Mistral needs to be proven to work reliably in multi-node mode. In order to achieve it we need to make a number of engine, executor and RPC changes and configure a CI gate to run stress tests on multi-node Mistral.

   Definition of done: CI gate supports MySQL, all critically important functionality (join, with-items, parallel workflows, sequential workflows) is covered by tests.

• Mistral Official Documentation
• Project status, bugs, and blueprints are tracked on Launchpad
• Additional resources are linked from the project Wiki page
• Apache License Version 2.0 http://www.apache.org/licenses/LICENSE-2.0