Drupal 8 Configuration - Extending the API

This is part four of a five-part series of articles exploring the Drupal 8 Configuration API, how it enables functionality to be migrated between multiple environments, and ways of using the Configuration API with contributed modules to effectively manage the configuration of a project.

15 May 2019

Background

We live in an age of Drupal complexity. In the early days of Drupal, many developers would have a single Drupal instance/environment (aka copy) that was their production site, where they would test out new modules and develop new functionality. Developing on the live website however sometimes met with disastrous consequences when things went wrong! Over time, technology on the web grew, and nowadays it's fairly standard to have a Drupal project running on multiple environments to allow site development to be run in parallel to a live website without causing disruptions. New functionality is developed first in isolated private copies of the website, put into a testing environment where it is approved by clients, and eventually merged into the live production site.

While multiple environments allow for site development without causing disruptions on the live production website, it introduces a new problem; how to ensure consistency between site copies so that they are all working with the correct code.

This series of articles will explore the Configuration API, how it enables functionality to be migrated between multiple environments (sites), and ways of using the Configuration API with contributed modules to effectively manage the configuration of a project. This series will consist of the following posts:

Part 1 gives the background of the Configuration API, as well as discusses some terminology used within this article, and Part 2 describes how the API works, and Part 3 explains how to use functionality provided by core, so they are worth a read before beginning this article. 

Read-only configuration

In some situations, site builders may want to prevent any configuration changes from being made on the production environment, preventing changes that may cause unexpected issues. For example, clients with admin access could log into the production server, and make what they think is an innocent configuration change, that results in unexpected and drastic consequences. Some site builders consider it to be a best practice to prevent configuration changes on the production server altogether, under the idea that only content should be editable on the production server, and configuration changes should only be made in development and/or staging environments before being tested and pushed to production.

The Config Readonly module, allows for configuration changes through the UI to be disabled on a given environment. It does this by disabling the submit buttons on configuration pages. The module also disables configuration changes using Drush and Drupal console.

Screenshot of the Configuration Readonly module output on configuration pages
A configuration form that has been disabled with the Configuration Readonly module

Note: some configuration forms may still be enabled when using this  module. Module developers must build their forms by extending ConfigFormBase for the Configuration Readonly module to do its magic. If the developer has built the form using other means, the form will not be disabled, and the configuration for that form can be changed through the admin UI.

To set up an environment as read-only, add the following line to settings.php, then enable the module:

$settings['config_readonly'] = TRUE;

After an environment is set as read-only, changes to configuration can only be made on other environments, then migrated and imported into the active configuration on the read-only environment.

Complete split (blacklist) configuration

Sometimes configuration needs to exist on some environments, but not exist in other environments. For example, development modules, like the Devel module, or UI modules like Views UI (Drupal core) and Menu UI (Drupal core) should not be enabled on production environments, as they add overhead to the server while being unnecessary since the production server should not be used for development.

A problem arises when configuration is exported from one environment, and imported into the production environment. All the configuration from the source environment is now the active configuration on the production environment. So any development modules that were enabled on the source environment are now enabled on the production environment. In the case of development modules like Devel, this may only add some overhead to the server, but imagine a module like the Shield module, which sets up environments to require a username and password before even accessing the site. If this module is accidentally enabled upon import on production, it will block the site from public access - a disaster!

The solution to this situation is to blacklist configuration. Blacklisted configuration is blacklisted (removed) from configuration upon export. This functionality is provided by the Configuration Split module. This module allows for black-listing configuration either by module, by individual configuration key(s), and/or by wildcard.

Note that more detailed directions for creating blacklists can be found on the documentation page. The following is meant to give an overview of how black lists work.

Blacklists are created as part of a configuration profile. Configuration profiles allow for 'splitting' (a divergence in) configuration between environments. Profiles may be created for environment types such development, staging and production allowing for configuration specific to those types of environments. Or profiles could be set up for public non-production environments, that would have the Shield module enabled and configured. While a development profile may apply to all development environments, not all development environments are on publicly accessible URLs, and therefore may not need the Shield module enabled.

When setting up a configuration profile, note that the folder name must be the same as the machine_name of the profile.

Screenshot of configuration split profile settings
Configuration split profile settings

Note that you must manually create the folder specified above, and that folder can and should be tracked using Git, so it can be use on any environment that enables the profile.

Configuration can then be set up to be blacklisted either by module, by configuration key, or by wildcard:

Screenshot of configuration split complete-split/blacklist settings
Complete split (blacklist) can be set by module, configuration key, or by wildcard

Finally, environments need to be set up to use a given profile. This is handled by adding the following line to settings.php on the environment:

$config['config_split.config_split.PROFILEMACHINENAME']['status'] = TRUE;

Where PROFILEMACHINENAME is the machine_name from the profile you created.

Although blacklisted configuration does not become part of the exported archive, it is not ignored altogether. When an environment has the profile enabled, upon export, blacklisted configuration is extracted, then written to the folder specified in the profile. The remaining configuration is written to the default configuration directory. When importing configuration, environments with the profile enabled will first retrieve the configuration from the default configuration directory, then apply any configuration from the folder specified in the profile. Environments not set up to use the profile ignore the configuration in the blacklisted directory altogether on both import and export.

This means that a developer can enable the Devel module on their local environment, blacklist it, then export their configuration. The blacklisted configuration never becomes part of the default configuration, and therefore the module will not accidentally be installed on environments with the configuration profile enabled.

Conditional split (grey list) configuration

Grey lists, also provided by the Configuration Split module, allow for configuration to differ by environment. With a blacklist (previous section), the configuration only exists in the active database configuration for environments that are set up to use the configuration profile containing the blacklisted configuration. With a grey list, the configuration exists in the active configuration in all environments, but the configuration profiles can be set up to allow environments to use differing values for the configuration.

Imagine an integration with a remote API requiring a username, password, and endpoint URL. The production server needs integrate with the remote API's production instance, while other environments will integrate with the remote API's sandbox instance. As such, the values to be used will differ by environment:

Production Environment:

remoteapi.username: ProductionUsername
remoteapi.password: ProductionPassword
remoteapi.endpoint: https://example.com/api

Other Environments:

remoteapi.username: SandboxUsername
remoteapi.password: SandboxPassword
remoteapi.endpoint: https://sandbox.example.com/api

A grey list allows for the setup of these values by configuration profile.

You may be remembering that Part 3 of this series of articles discussed overriding configuration in settings.php, and thinking that a grey list sounds like the same thing. After all, the default values for the sandbox instance of the API could be set up as the configuration values, and the production values could be overridden in settings.php on the production environment, with the same end-result.

The difference is that with a grey list, the remote API values are saved to the configuration profile folder, which is tracked by Git, and therefore can be tracked and migrated between environments. When grey listed configuration is exported, the grey listed configuration is written to the configuration profile folder, in the same manner as blacklisted configuration. When configuration is imported, the default values are retrieved, and the grey list values are used to override the default values, after which the configuration is imported into active configuration.

With the configuration override method using settings.php, site builders need to store the various configuration values somewhere outside the project, communicating environment-specific configuration values to each other through some means, to be manually entered on the relevant environment(s). With a grey list, the configuration values are managed with Git, meaning site builders do not need to record them outside the project, nor communicate them to each other through some other means. Site builders simply need to enable the relevant configuration profile in settings.php, and the environment-specific values can then be imported into active configuration from the configuration profile directory. This means that the sandbox API values can be set up as the values used by default on all environments, and a production configuration profile can be enabled on the production environment using the values to connect to the production instance of the remote API.

Conditional split items can be selected either from a list, or by manually entering them into the configuration profile:

Screenshot of configuration split conditional-split/grey list settings
Conditional split (grey list) settings can be selected or manually entered

Finally, note that grey lists can actually be used in conjunction with configuration overrides in settings.php. Grey lists are applied during import and export of configuration from the database. Values in settings.php are used at runtime, overriding any active configuration. So a developer could choose to set up their local instance of the system to connect to an entirely different instance of the remote API altogether by overriding the values in settings.php.

Ignoring configuration (overwrite protection)

Sometimes developers will want to protect certain configuration items in the database from ever being overwritten. For example imagine a site named Awesome Site, with a module that supplies the core of the site, named awesome_core. Since this module provides the core functionality of the site, it should never be disabled under any circumstances, as that would disable the core functionality of the site. In this case, the configuration for this module can be set to be 'ignored'. Any attempts to import ignored configuration from the file system to the active configuration in database will be skipped, and not imported.

Configuration can be ignored using the Config Ignore module. The functionality this module provides is similar to the functionality provided by the Config Readonly module discussed earlier, however the Config Readonly module covers the entire configuration of an environment, while the Config Ignore module allows for choosing configuration that should be protected. This configuration is protected by ignoring it altogether on import.

Configuration can be ignored as follows:

  1. Enable Config Ignore module on all environments.
  2. Navigate to the config ignore UI page, and set the configuration item to be ignored. In the case of preventing the awesome_core module from being disabled, the following would be added:
    core.extension:module.awesome_core
    Screenshot of the Configuration Ignore module admin page
    Configuration to be ignore is entered one item per line. Wildcards can be used.

This setting will ensure that any attempts to change or remove core.extension:module.awesome_core upon configuration import will be ignored. So if the module is enabled on production, and a developer pushes configuration changes that would uninstall this module, those changes will be ignored, and the module will still be set as enabled after import.

Summary

In this article, we looked at various modules that extend the Configuration API, use cases behind these modules, and how the modules worked. We looked at the Config Readonly module, the Configuration Split module, and the Config Ignore module, and how to use these modules to manage configuration differences between environments. In the next, final fifth part of this series, we will look at configuration management for module developers, and how developers can define the schema for the configuration in modules they develop.

 

More like this