CHANGES.rst

Tequila-django

Changes

v 0.9.28 on Feb 24, 2020

• Remove support for gunicorn_version. Instead, projects should pin the specific gunicorn version that they want to use in their requirements.txt file.

v 0.9.27 on Oct 17, 2019

• Allow overriding the default supervisor version.

v 0.9.26 on Oct 2, 2019

• Omit version parameter to "optionally install newrelic" task when new_relic_version is an empty string.
• Change file permissions for supervisor, to allow group access.

v 0.9.25 on Sep 24, 2019

• Consistently use the recommended module:options format for all tasks.
• Change include to include_tasks.
• Add app_packages variable for specifying additional system packages to be installed.
• Add support for using uWSGI instead of gunicorn.

v 0.9.24 on Jul 5, 2019

• Add variable force_recreate_venv to allow virtualenv to be deleted, forcing it to be refreshed from scratch.

v 0.9.23 on Apr 8, 2019

• Don't follow symlinks when changing ownership of copied source code, since they might point to things on the dev's system that don't make sense on the server and even result in infinite recursion trying to follow them.

v 0.9.22 on Apr 5, 2019

• Allow pinning to a specific version of New Relic with new_relic_version

v 0.9.21 on Jan 28, 2019

• When rsyncing local source to server (when source_is_local is true, mainly when using vagrant), exclude deployment dir, and use the same ssh as ansible.

v 0.9.20 on Dec 27, 2018

• Remove --log-syslog from the supervisor gunicorn command line. Supervisor was already logging the gunicorn stdout to syslog and trying to log it twice was failing.

v 0.9.19 on Nov 21, 2018

• Add ability to customize Celery app name via celery_app variable.

v 0.9.18 on Sep 26, 2018

• Allow a celery beat instance to be deployed and configured independently of the celery worker instances. This is because it is desirable to have only one beat instance, generally.

  It is recommended to add a new group to your inventory files, e.g. [beat], that has exactly one instance in it, which may be an instance that is also in your [worker] group. You may then create a new playbook, e.g. deployment/playbooks/beat.yml, containing something like:

  ---
  - hosts: beat
    become: yes
    roles:
      - { role: tequila-django, is_celery_beat: true }
  

  If you go this route, do not forget to add this new beat.yml file to the list invoked by your site.yml playbook. Note that if you incorporate this change this way, the tequila-django role will be executed multiple times on instances that incorporate these differing variants. Alternatively, one can fold together the invocation of tequila-django into a single playbook that uses group checking to set the parameters used, like so:

  ---
  - hosts: web:worker:beat
    become: yes
    roles:
      - role: tequila-django
        is_web: "{{ 'web' in group_names }}"
        is_worker: "{{ 'worker' in group_names }}"
        is_celery_beat: "{{ 'beat' in group_names }}"
  

  This has the advantage of requiring only one loop through the tequila-django tasks per server instance.

v 0.9.17 on Sep 11, 2018

• Consistently use manage.py and dotenv.sh to invoke Django with the same environment.

v 0.9.16 on Aug 6, 2018

• New variable project_subdir to accomodate projects where the Django project (manage.py etc.) are in a subdirectory of the repository.
• New variable wsgi_module to accomodate projects where the package of the wsgi module is not {{ project_name }}.wsgi.

v 0.9.15 on July 31, 2018

• Fix syntax error in celery.yml

v 0.9.14 on July 30, 2018

• New variable project_port to have Django listen on a port other than 8000, to help with multi-site deploys.
• Update README with example for SSH agent forwarding.

v 0.9.13 on June 1, 2018

• Add additional variables to enable Celery event monitoring with snapshots.

  Per the remarks added to the "Variables" section of the README, tequila-django now includes variables that allow the user to turn on Celery event monitoring alongside other Celery commands. This relies on the celery events command introduced in Celery 3.1, and by default it uses the django-celery-monitor app to capture worker events.

v 0.9.12 on May 31, 2018

• Split out the npm tasks into a new role, tequila-nodejs. Projects upgrading to this version will need to add tequila-nodejs to their requirements.yml file while still retaining geerlingguy.nodejs, which is a dependency of this new role. Additionally, the playbook/web.yml playbook should be updated to execute tequila-nodejs instead of geerlingguy.nodejs, and it should be placed immediately after the tequila-django role. See the new example web.yml file in the playbooks directory in the caktus/tequila repo.

v 0.9.11 on March 19, 2018

• Add additional env vars to account for Celery 4's renamed configuration settings.

  NOTE: if you are upgrading from Celery 3 to Celery 4, be aware of the changes made to Celery's setting names. Most relevant to tequila-django, the changed names include the environment-variable-dependent settings determined by the secrets broker_host and broker_password.

  Old setting names will still work after the change, but users are encouraged to upgrade as soon as possible. Settings can be upgraded automatically using the Celery 4 command line interface.

  For more details, see the Celery 4 "What’s new in Celery 4.0" changelog document, in particular the "For Django users and others who want to keep uppercase names" section.

v 0.9.10 on Mar 2, 2018

• Move celerybeat's pid file to an in-memory filesystem.

v 0.9.9 on Dec 12, 2017

• Throw a failure if the presence of packages in the project package.json devDependencies object is detected. Projects will need to move their dependencies into the dependencies object instead, or disable the check by setting ignore_devdependencies to true.

v 0.9.8 on Nov 27, 2017

• Properly quoting all environment variables.

  Previously, most of the Ansible variables that were dropped into the .env file were not being quoted within that file, making many characters (spaces, characters with special meaning to the shell, etc.) unsafe for use within this file. The workaround was to double-quote your Ansible variables, but now with this change any such variables need to be identified and the extra quoting removed.

v 0.9.7 on Oct 19, 2017

• Work around for Ansible bug #25431. Store Github deploy key in a fact so it's available regardless of result of previous task.

v 0.9.6 on Sep 21, 2017

• Added an additional_domains variable, to allow for the multi-domain case to be supported in the Django settings file.

v 0.9.5 on Sep 13, 2017

• Added a cloud_staticfiles boolean variable, that determines whether to run collectstatic only once or not.

v 0.9.4 on Sep 6, 2017

• Mark the Django migration task to only run once.

v 0.9.3 on Aug 28, 2017

• Allow Git checkout of source tree on Vagrant environments instead of only supporting source_is_local=true.
• Ignore .env and node_modules when rsync-ing source tree for source_is_local processing.

v 0.9.2 on Aug 17, 2017

• Adjustment to the Django migrate task to improve its speed, and a suggestion for a project-wide setting to improve it further.

v 0.9.1 on July 20, 2017

• Convert the collectstatic task into a handler. This is needed so that it happens after the npm run build step, so that the files generated from that are included.

v 0.9.0 on July 18, 2017

• Remove the nodejs installation and package management in favor of the geerlingguy/nodejs Ansible role.

  Important

  To upgrade to this version, you will have to make the following changes to your deployment files.

  1. Add the geerlingguy/nodejs role to deployment/requirements.yml, and bump the version of tequila-django:

     ---
     - src: https://github.com/caktus/tequila-django
       version: v0.9.3
       name: tequila-django
     
     - src: geerlingguy.nodejs
       version: 4.1.2
       name: nodejs
     ...
     

  2. Install the new role, and make sure that tequila-django gets upgraded. Since ansible-galaxy does not at this time seem to have support for version upgrades, either explicitly remove the tequila-django directory from deployment/roles/, or use ansible-galaxy uninstall tequila-django, before running the command to install the roles from the requirements.yml file.

  3. Include the configuration variables for geerlingguy/nodejs in your project-wide variables file (usually deployment/playbooks/group_vars/all/project.yml):

     ---
     nodejs_version: "6.x"
     nodejs_install_npm_user: "{{ project_name }}"
     nodejs_package_json_path: "{{ source_dir }}"
     nodejs_config_unsafe_perm: true
     

     If you previously had anything configured under the variable global_npm_installs, rename this variable to nodejs_npm_global_packages. Note that nodejs_config_unsafe_perm has to be set to true in order for the global npm installs to work for nodejs_install_npm_user set to anything other than root.

  4. If you previously had a different version of nodejs installed using a .deb package, you should probably uninstall it, and remove any PPA source file associated with it (if used).

  5. Modify your deployment/playbooks/web.yml file (or equivalent) to include the nodejs role _after_ the tequila-django role:

     ---
     - hosts: web
       become: yes
       roles:
         - tequila-nginx
         - { role: tequila-django, is_web: true }
         - nodejs