Install and configure for Ubuntu

This section describes how to install and configure the Workflow Service service for Ubuntu.


  1. Install the packages:

    # apt-get update
    # apt-get install python-setuptools python-pip libffi-dev libxslt1-dev \
      libxml2-dev libyaml-dev libssl-dev python3-dev tox mistral-common


NOTE: For instructions on how to install Mistral using devstack, refer to Mistral Devstack Installation

Clone the repo and go to the repo directory:

$ git clone
$ cd mistral

Generate the configuration file:

$ tox -egenconfig

Create the mistral directory and copy the example configuration file:

$ mkdir /etc/mistral
$ cp etc/mistral.conf.sample /etc/mistral/mistral.conf

Edit the configuration file:

$ vi /etc/mistral/mistral.conf

Virtualenv installation:

$ tox

This installs the necessary virtual environments and runs all the project tests. Installing the virtual environments may take significant time (~10-15 mins).

Local installation:

$ pip install -e .


$ pip install -r requirements.txt
$ python install

NOTE: There are some differences between pip install -e and install. pip install -e works similarly to install or the EasyInstall tool, however, it does not actually install anything. Instead, it creates a special .egg-link file in the deployment directory that links to your project’s source code.

Mistral Configuration and Policy Guide

Before the first run

After the installation, you will see the mistral-server and mistral-db-manage commands in your environment, either in system or virtual environment.

NOTE: If you use virtualenv, all Mistral-related commands can be accessed with tox -evenv –. For example, mistral-server is available via tox -evenv – mistral-server.

The mistral-db-manage command can be used for migrations.

Updating the database to the latest revision type:

$ mistral-db-manage --config-file <path-to-mistral.conf> upgrade head

Before starting the Mistral server, run the mistral-db-manage populate command. It creates the DB with all the standard actions and standard workflows that Mistral provides to all Mistral users.:

$ mistral-db-manage --config-file <path-to-mistral.conf> populate

For more detailed information on the mistral-db-manage script, see the Mistral Upgrade Guide.

NOTE: For users who want a dry run with an SQLite database backend (not used in production), the mistral-db-manage script is not recommended for database initialization because of SQLite limitations. Use the sync_db script described below for database initialization instead.

If you use virtualenv:

$ tools/ --config-file <path-to-mistral.conf>

Or run sync_db directly:

$ python tools/ --config-file <path-to-mistral.conf>

Running Mistral API server

To run the Mistral API server, execute the following command in a shell:

$ mistral-server --server api --config-file <path-to-mistral.conf>

Running Mistral Engines

To run the Mistral Engine, execute the following command in a shell:

$ mistral-server --server engine --config-file <path-to-mistral.conf>

Running Mistral Executors

To run the Mistral Executor instance, execute the following command in a shell:

$ mistral-server --server executor --config-file <path-to-mistral.conf>

Note that at least one Engine instance and one Executor instance should be running so that workflow tasks are processed by Mistral.

Mistral Notifier

To run the Mistral Notifier, execute the following command in a shell:

$ mistral-server --server notifier --config-file <path-to-mistral.conf>

Running Multiple Mistral Servers Under the Same Process

To run more than one server (API, Engine, or Task Executor) on the same process, execute the following command in a shell:

$ mistral-server --server api,engine --config-file <path-to-mistral.conf>

The –server command line option can be a comma delimited list. The valid options are “all” (by default if not specified) or any combination of “api”, “engine”, and “executor”. It is important to note that the “fake” transport for the rpc_backend defined in the config file should only be used if “all” the Mistral servers are launched on the same process. Otherwise, messages do not get delivered if the Mistral servers are launched on different processes because the “fake” transport is using an in-process queue.

Mistral Client Installation Guide

To install python-mistralclient, it is required to have pip (in most cases). Make sure that pip is installed. Then type:

$ pip install python-mistralclient

Or, if it is needed to install python-mistralclient from master branch, type:

$ pip install git+

After python-mistralclient is installed you will see command mistral in your environment.

Configure authentication against Keystone

If Keystone is used for authentication in Mistral, then the environment should have auth variables:

$ export OS_AUTH_URL=http://<Keystone_host>:5000/v2.0
$ export OS_TENANT_NAME=tenant
$ export OS_USERNAME=admin
$ export OS_PASSWORD=secret
$ export OS_MISTRAL_URL=http://<Mistral host>:8989/v2
  ( optional, by default URL=http://localhost:8989/v2)

and in the case when you are authenticating against keystone over https:

$ export OS_CACERT=<path_to_ca_cert>


In client, we can use both Keystone auth versions - v2.0 and v3. But server supports only v3.

You can see the list of available commands by typing:

$ mistral --help

To make sure Mistral client works, type:

$ mistral workbook-list

Configure authentication against Keycloak

Mistral also supports authentication against Keycloak server via OpenID Connect protocol. In order to use it on the client side the environment should look as follows:

$ export MISTRAL_AUTH_TYPE=keycloak-oidc
$ export OS_AUTH_URL=https://<Keycloak-server-host>:<Keycloak-server-port>/auth
$ export OS_TENANT_NAME=my_keycloak_realm
$ export OS_USERNAME=admin
$ export OS_PASSWORD=secret
$ export OPENID_CLIENT_ID=my_keycloak_client
$ export OPENID_CLIENT_SECRET=my_keycloak_client_secret
$ export OS_MISTRAL_URL=http://<Mistral host>:8989/v2
 (optional, by default URL=http://localhost:8989/v2)


Variables OS_TENANT_NAME, OS_USERNAME, OS_PASSWORD are used for both Keystone and Keycloak authentication. OS_TENANT_NAME in case of Keycloak needs to correspond a Keycloak realm. Unlike Keystone, Keycloak requires to register a client that access some resources (Mistral server in our case) protected by Keycloak in advance. For this reason, OPENID_CLIENT_ID and OPENID_CLIENT_SECRET variables should be assigned with correct values as registered in Keycloak.

Similar to Keystone OS_CACERT variable can also be added to provide a certification for SSL/TLS verification:

$ export OS_CACERT=<path_to_ca_cert>

In order to disable SSL/TLS certificate verification MISTRALCLIENT_INSECURE variable needs to be set to True:


Targeting non-preconfigured clouds ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^-

Mistral is capable of executing workflows on external OpenStack clouds, different from the one defined in the mistral.conf file in the keystone_authtoken section. (More detail in the Mistral Configuration and Policy Guide).

For example, if the mistral server is configured to authenticate with the cloud and the user wants to execute the workflow on the cloud.

The mistral.conf will look like:

www_authenticate_uri =

The client side parameters will be:

$ export OS_AUTH_URL=
$ export OS_USERNAME=mistral_user
$ export OS_TARGET_USERNAME=cloud_user


Every OS_* parameter has an OS_TARGET_* correspondent. For more detail, check out mistral –help

The OS_* parameters are used to authenticate and authorize the user with Mistral, that is, to check if the user is allowed to utilize the Mistral service. Whereas the OS_TARGET_* parameters are used to define the user that executes the workflow on the external cloud,

Use cases

Authenticate in Mistral and execute OpenStack actions with different users

As a user of Mistral, I want to execute a workflow with a different user on the cloud.

Execute workflows on any OpenStack cloud

As a user of Mistral, I want to execute a workflow on a cloud of my choice.

Special cases

Using Mistral with zero OpenStack configuration:

With the targeting feature, it is possible to execute a workflow on any arbitrary cloud without additional configuration on the Mistral server side. If authentication is turned off in the Mistral server (Pecan’s auth_enable = False option in mistral.conf), there is no need to set the keystone_authtoken section. It is possible to have Mistral use an external OpenStack cloud even when it isn’t deployed in an OpenStack environment (i.e. no Keystone integration).

With this setup, the following call will return the heat stack list:

$ mistral \
    --os-target-auth-url= \
    --os-target-username=testuser \
    --os-target-tenant=testtenant \
    --os-target-password="MistralRuleZ" \
    run-action heat.stacks_list

This setup is particularly useful when Mistral is used in standalone mode, when the Mistral service is not part of the OpenStack cloud and runs separately.

Note that only the OS-TARGET-* parameters enable this operation.

Finalize installation

Restart the Workflow services:

# service openstack-mistral-api restart