authelia/docs/deployment-dev.md

5.2 KiB

Deployment for Dev

  1. Deploy With npm
  2. Deploy With Docker
  3. Deploy nginx
  4. Discard components
    1. Discard MongoDB
    2. Discard Redis
    3. Discard LDAP
  5. FAQ

WARNING: the instructions given in this documentation are not meant to be used for production environments since it will make Authelia non resilient to failures.

NOTE: If not done already, we highly recommend you first follow the [Getting Started] documentation.

In some cases, like protecting personal websites, it can be fine to use Authelia in a non highly-available manner. Fortunately, we can achieve to run it along with one reverse proxy meaning the setup is then reduced to only two components: Authelia and nginx.

As for a regular deployment in production, you need to install Authelia either by pulling the Docker image or installing the npm package and run it with a configuration file passed as argument.

Deploy with npm

npm install -g authelia
authelia /path/to/your/config.yml

Deploy with Docker

docker pull clems4ever/authelia
docker run -v /path/to/your/config.yml:/etc/authelia/config.yml clems4ever/authelia

Deploy Nginx

You also need to install nginx and take example/compose/nginx/minimal/nginx.conf as an example of configuration.

Discard components

Discard MongoDB

There is an option in the configuration file to discard MongoDB and use your local filesystem to store the database data. This option will therefore prevent you from running multiple instances of Authelia in parallel. Consequently, this option is not meant to be used in production.

Here is the configuration block you should use:

storage:
  # The directory where the DB files will be saved
  local:
    path: /var/lib/authelia/store

Discard Redis

There is an option in the configuration file to discard Redis and use the memory of the server to store the KV data. This option will therefore prevent you from running multiple instances of Authelia in parallel and will make you lose user sessions if the application restarts. This concretely means that all your users will need to authenticate again in that case. Hence, this option is not meant to be used in production.

To use memory instead of a Redis backend, just comment out the Redis connection details in the following block:

session:
  ...      
  # # The redis connection details
  # redis:
  #  host: redis
  #  port: 6379
  #  password: authelia

Discard LDAP

Authelia can use a file backend in order to store users instead of a LDAP server or an Active Directory. This mode will therefore prevent you from running multiple instances of Authelia in parallel and is therefore discouraged for production environments.

To use a file backend instead of a LDAP server, you should first duplicate the file users_database.yml and edit it to add the users you want.

The content of this file is as follows:

users:
  ...
  john:
    password: "{CRYPT}$6$rounds=500000$jgiCMRyGXzoqpxS3$w2pJeZnnH8bwW3zzvoMWtTRfQYsHbWbD/hquuQ5vUeIyl9gdwBIt6RWk2S6afBA0DPakbeWgD/4SZPiS0hYtU/"
    email: john.doe@authelia.com
    groups:
      - admins
      - dev

The password is hashed and salted as it is in LDAP servers with salted SHA-512. Here is a one-liner to generate such hashed password:

python3 -c 'import crypt; print("{CRYPT}" + crypt.crypt("mypassword", crypt.mksalt(crypt.METHOD_SHA512, rounds=500000)))'

Once the file is created, edit the configuration file with the following block (as used in config.minimal.yml):

authentication_backend:
  file:
    path: /etc/authelia/users_database.yml

instead of (used in config.template.yml):

authentication_backend:
  ldap:
    url: ldap://openldap
    base_dn: dc=example,dc=com

    additional_users_dn: ou=users
    users_filter: cn={0}

    additional_groups_dn: ou=groups
    groups_filter: (&(member={dn})(objectclass=groupOfNames))

    group_name_attribute: cn

    mail_attribute: mail

    user: cn=admin,dc=example,dc=com
    password: password

FAQ

Can you give more details on why this is not suitable for production environments?

This documentation gives instructions that will make Authelia non highly-available and non scalable by preventing you from running multiple instances of the application. This means that Authelia won't be able to distribute the load accross multiple servers and it will prevent failover in case of a crash or an hardware issue. Moreover, it will also prevent from reliably persisting data and consequently fail access to your platform as the devices registered by your users will be lost.

Why is this not automated?

Well, as stated before those instructions are not meant to be applied for a production environment. That being said, in some cases it is just fine and so writing an Ansible playbook to automate all this process would be great. We would really be more than happy to review such a PR.