Joplin - CMS for the City of Austin

Joplin is the Authoring Interface for adding and editing content for alpha.austin.gov.

The public facing frontend (Janis) displays the content that is published in Joplin.

Joplin is built using Wagtail, a Content Management System (CMS) using Python's django framework.

Index

• How to Run Locally
• Testing
• Updating the Data Model
• CircleCI Deployments
• Useful Commands
• Performance Monitoring
• Debugging
• Design
• Related Repos
• Misc

---

How to Run Locally

First, install docker (version 18.09 or greater) and clone this repo.

Add .env file

Copy the template and modify for your local environment as you see fit.

cp template.env .env

This will automatically load environment variables into your pipenv environment. (If you choose to run Joplin on your host machine without docker). Commands like pipenv run ./joplin/manage.py migrate will then know which database to use.

Run without data

./scripts/serve-local.sh

• This will get you started with one admin user and no data.
• It will automatically run all django migrations.
• It will start up 3 docker containers: joplin_app_1 (for running the CMS web server), joplin_assets_1 (for managing assets), and joplin_db_1 (for the postgres database).
• Viewing your docker logs, you can tell that your server is running successfully when you see these listeners:
• Your Joplin instance will be accessible at http://localhost:8000/admin with the credentials user: admin@austintexas.io, pw: x

Or if you prefer to run without docker (for speed + ability to integrate with debugging tools):

./scripts/undockered.sh

Run with prod data

LOAD_DATA="prod" ./scripts/serve-local.sh
LOAD_DATA="prod" ./scripts/undockered.sh

• This will add some seeding content from the last prod datadump (joplin/db/system-generated/prod.datadump.json) created by migration-test.sh.
• LOAD_DATA and LOAD_PROD_DATA both load from the prod.datadump.json
• You have to run the LOAD_x_DATA commands on a clean db instance. You can wipe the DB and load data in the same step by running:
  • DROP_DB=on LOAD_DATA="prod" ./scripts/serve-local.sh

Run with staging data

LOAD_DATA="staging" ./scripts/serve-local.sh
LOAD_DATA="staging" ./scripts/undockered.sh

• This will add some seeding content from the last staging datadump (joplin/db/system-generated/staging.datadump.json) created by migration-test.sh.

Run with dummy data

LOAD_DATA="dummy" ./scripts/serve-local.sh
LOAD_DATA="dummy" ./scripts/undockered.sh

• This will add dummy content from the last dummy datadump (joplin/db/system-generated/dummy.datadump.json) created by migration-test.sh.

Drop Existing DB

DROP_DB=on ./scripts/serve-local.sh

Drop DB, run with fresh data

RELOAD_DATA=$SOURCE ./scripts/serve-local.sh

• shorthand for LOAD_DATA=$SOURCE + DROP_DB=on

Run with Janis

JANIS="on" ./scripts/serve-local.sh

• Runs with the "janis:local" image built on your machine.
• Can be combined with any other args.

Run + Rebuild without cached image layer

If something goes wrong with your docker builds and you want to start over without any cached layers, you can run:

HARD_REBUILD="on" ./scripts/serve-local.sh

• LOAD_DATA=$SOURCE can also be used with HARD_REBUILD="on"
• It takes 90 seconds to do a HARD_REBUILD.
• If worse comes to worse, you can always delete your local joplin docker images with docker rmi.

Run undockerized

You might prefer to run the Django app on your host computer to enable better access to debugging tools. This script will still run joplin_assets and joplin_db on docker containers, but will run a django runserver command directly on your host computer.

All of the above flags (such as LOAD_DATA=on) will work with the undockered version of Joplin.

./scripts/undockered.sh

If you run into pipenv errors or are running this for the first time, you can build/rebuild a pipenv with:

REBUILD_PIPENV=on ./undockered.sh

Override default behavior of stopping existing Joplin containers

NO_STOP=on ./scripts/undockered.sh

• Makes undockered development go a little faster. You don't need to turn off then turn on the helper DB and Assets containers.

Run with custom smuggler data (*Possibly deprecated)

If you don't want to load the default data used in LOAD_DATA=$SOURCE, you have to ability to source data from any environment you'd like using a django plugin called smuggler.

To load in data from smuggler follow these steps:

1. Download a json datadump from the Joplin deployment of your choosing by visiting [joplin URL]/django-admin/dump.
2. Place your datadump in the smuggler fixtures directory joplin/db/smuggler
3. Start a data-less empty local Joplin instance with ./scripts/serve-local.sh.
   • Note! If you previously loaded data locally, make sure to delete your existing joplin_db_1 container before this step.
4. Go to your local Joplin's smuggler interface at http://localhost:8000/django-admin/load/. Select the data file that you want to load.
5. At this point the database should be populated, but any media used on the site will be broken, to fix this run: ./scripts/download-media.sh

Access Local DB

psql postgres://joplin@localhost:5433/joplin

The database defaults to version 10 of postgres. No password is set up, since there is no security needed for a local environment. To connect, use the localhost at the standard PostgreSQL (5432) port where the container is mapped to listen for connections all without a password, the user and database name is 'joplin'. To manage the database, you should be able to use your favorite DB admin tool (ie. TablePlus, DBeaver, DataGrip, etc).

It runs on PORT 5433 to avoid port conflicts with your host's default postgres port.

Note: The containers are not built at the same time; for this purpose, joplin will wait and display a 'database not available' message in a loop until the database is up and ready. This is because the DB container takes a little longer to build and set up locally, and joplin has to wait before it can run the django migrations locally.

Handling schema migration conflict

Sometimes, you may run ./scripts/serve-local.sh and find that the joplin container cannot run due to a migration conflict.

If you check the log, you might see an error like this:

app_1     | CommandError: Conflicting migrations detected; multiple leaf nodes in the migration graph: (0073_auto_20190604_2124, 0069_auto_20190530_2220 in base).
app_1     | To fix them run 'python manage.py makemigrations --merge'

In that case, to fix the error run:

./scripts/run-handle-migration-conflicts.sh

This will run makemigrations with the --merge flag, and should do a good job handling simple conflicts.

See more: https://docs.djangoproject.com/en/2.2/topics/migrations/

---

Testing

Testing documentation can be found here.

---

Updating the Data Model

1. Have a local Joplin instance running (probably populated with data).
2. Update your data model wherever you have it written.
3. Make a new migration with:
   • docker exec -it joplin_app_1 python joplin/manage.py makemigrations
4. Run that migration with:
   • docker exec -it joplin_app_1 python joplin/manage.py migrate
5. Test that your migration works with
   • SOURCE=prod ./scripts/migration-test.sh

Adding new content types

Adding new content types to Joplin is a fairly involved process, and there are many things need to go from a new data model to new content appearing in Janis. We won't cover all that here, but just some basic pointers.

Django is fairly flexible when it comes to how you organize your code. However, a typical and often recommended approach is to break components into individual apps.

You can do this a couple ways, but a simple way using the 'startapp' command, like so:

python manage.py startapp coolnewcontenttype

You can then add that to the list of installed apps in joplin's settings files. More info here: https://docs.djangoproject.com/en/2.2/ref/django-admin/#startapp Wise words here: https://docs.djangoproject.com/en/3.0/intro/reusable-apps/

Once you've got the new content type modeled, you can also expose the model as needed for our GraphQL api by adding the appropriate code to the 'api' application. There are several pre-existing patterns there that can get you started.

About migration-test script

The migration-test script makes sure that your migration changes will work even when they are applied to a database running the last migration. This is basically a dry run of a merge to the master branch of Joplin. If they do work, then the script will create a new datadump (to be used by LOAD_DATA="prod") with the new migrations applied. This will prevent future datadump schema version conflicts (which will happen if your datadump is from a different migration version than the Joplin instance its going into).

Note: This process does not update staging. It updates the data that is seeded into local and PR builds. Staging data is persistent by design and would need to be manually updated.

Options:

• "SOURCE=prod sh scripts/migration-test.sh"
  • Sources data from production database
  • Builds migrations from "cityofaustin/joplin-app:production-latest" image
  • Then applies your local migrations on top of that
• "SOURCE=prod USE_PRIOR_DATADUMP=on sh scripts/migration-test.sh"
  • Sources data from your existing prod.datadump.json
• "SOURCE=staging sh scripts/migration-test.sh"
  • Sources data from staging database
• "SOURCE=staging USE_PRIOR_DATADUMP=on sh scripts/migration-test.sh"
  • Sources data from your existing prod.staging.json

Bonus extra optional Params:

• "DOCKER_TAG_DB_BUILD=[x]" will build initial migrations from the docker image of your choice. Potentially could be used if you intend to merge into a branch other than master.
• "JANIS=on" will automatically spin up a Janis container for you. Note: you must have a "janis:local" image available locally.

Here's what migration-test.sh does at a high level:

1. Creates a database from the last working Joplin migration
   • This is accomplished by running Joplin migrations and data loading from the joplin_app image tagged as cityofaustin/joplin_app:master-latest from the City of Austin's dockerhub.
   • Note: if you aren't intending to fork from/merge into the latest master branch, you can manually plug in your own DOCKER_TAG_DB_BUILD to test how you migrations work against a different build by running DOCKER_TAG_DB_BUILD=[joplin-app:some-tag] migration-test.sh.
2. Runs your new migrations on the old database
   • The previous joplin_app container shuts down (but the joplin_db stays up). Now a new joplin_app container (built from your local Joplin directory, tagged as "joplin_app:local") runs against the old joplin_db. The new migrations are automatically applied through joplin_app's entrypoint.
3. Spins up a local Janis and Joplin for you to test manually.
   • If you pass JANIS=on ./scripts/migration-test.sh then it will automatically spin up a Janis image using your own janis:local image. Otherwise, at this step you can manually start a Janis instance using another method.
   • Make sure that Joplin and Janis work as expected and that nothing breaks on Janis.
   • A command line prompt will ask if the migration worked. If you enter "y", then a new datadump fixture will replace the old seeing.datadump.json fixture in joplin/db/system-generated. If you enter "n", then the migration_test containers will shut down and not replace your datadump fixture.

Updating Dummy Data

Running SOURCE=dummy ./scripts/migration-test.sh will load in the latest dummy datadump and run migration test in dummy data mode. I (Brian) have been running this and then adding data when it gets to the interactive step. Once I'm happy with the data I have I respond to the Is it all good? question with y and get a shiny new dummy.datadump.json.

CircleCI Deployments

We use CircleCI to manage and customize our deployments to our cloud infrastructure. Important files to consider:

.circleci/config.yml

This file contains the stages and commands to execute, and the order of execution. It also contains the conditions under which a deployment can happen.

/circleci/docker

The contains the docker images used during circleci builds. The joplin-ci image is the container that runs our .circleci/config.yml process within circleci. joplin-base is the foundational image for all joplin builds. It speeds up .circleci deployment time to have it pre-built and stored in dockerhub (rather than as part of joplin-common within app.Dockerfile).

Builds for these images are done manually as needed and then stored in the cityofaustin dockerhub repo:

sh .circleci/docker/push.sh [name of Dockerfile]

After your updated image is pushed, update the image tags in /circleci/config.yml to use your new git commit SHA tag.

Steps

CircleCI will run the deployment workflow for every branch. Certain branches will have different steps applied to them.

The deployment is handled in four different stages, for each job circleci will create a new container (with a temporal (ephemeral) file system) and run all the operations as indicated per stage.

1. test

.circleci/scripts/cli_test.sh

At the moment the pre-deployment tests only check whether the AWS & Heroku CLI tools have been installed and are properly running.

2. backup_database

.circleci/scripts/backup_database.sh

If deploying on master or production, we first take a backup of the database and store it in S3 for disaster recovery. The backup file name is comprised of these values:

• The name of the application
• The timestamp
• The SHA of the latest commit
• The name of the latest django migration

Once the name is generated then it proceeds to generate a full URL link where the final file will be stored in S3. It connects to the database and generates a full backup and automatically saves it to S3 using the established nomenclature.

3. build_heroku_infrastructure

.circleci/scripts/build_heroku_infrastructure.sh

Builds infrastructure for a new heroku app. Rebuilds database for an existing heroku app if the database does not exist. This step is not done on the production branch because production should already have a heroku app and database in place.

create_review_app.sh is where your environment variables get set in heroku for a new app.

4. set_pr_vars

.circleci/scripts/set_pr_vars.sh

Adds environment variables to PR apps. Environment vars for staging and production are handled manually within Heroku console.

set_pr_vars.sh is where environment variables get updated after your app has been created.

.circleci/vars/branch_overrides.py is where you can set environment variables for one specific branch.

5. build_and_release

.circleci/scripts/build_image.sh Builds the Joplin docker image and pushes to cityofaustin's dockerhub repo and the heroku app.

.circleci/scripts/release.sh

Releases the docker image pushed to heroku in the previous step. This is what actually deploys the image to heroku.

.circleci/scripts/migrate.sh

Heroku Docker entrypoints time out at 60 seconds. So django migrations and data loading need to be initiated by invoking the docker-entrypoint.sh script manually.

The migration process currently consists of 3 commands:

• python ./joplin/manage.py migrate - This command will trigger the django migration process. It runs for all instances on the Cloud: PRs, Staging and Production.
• python ./joplin/manage.py loaddata - This command will import the latest datadump in joplin/db/system-generated. This only happens for PR branches.
• python ./joplin/manage.py collectstatic - Triggers the collect static process which only takes care of wagtail and certain images. This only runs for staging and production.

---

Useful Commands

• ssh into a container that's deployed on Heroku

  • heroku run bash -a $APPNAME

• Copy data from one app into another

  • heroku pg:copy source-app::DATABASE_URL DATABASE_URL -a destination-app
  • Plug in the actual names of your "source-app" and "destination-app".
  • Literally use "DATABASE_URL" in your command.

• Shut down all joplin containers:

  • source scripts/docker-helpers.sh; stop_project_containers joplin

• Delete all joplin containers:

  • source scripts/docker-helpers.sh; delete_project_containers joplin

• Create New App:

  • APP_NAME=app_name_goes_here
    docker exec joplin /bin/bash -c "mkdir -p \"$APP_NAME\" && cd joplin && python manage.py startapp \"$APP_NAME\""
    

• Access the Graphql API

  • localhost:8000/api/graphiql
  • Once you enter your username and password, you will be able to query the graphiql endpoint.

• Troubleshooting:

  • Clean up older docker images and containers and rebuild the new application if necessary:

  • # Assuming you only have joplin containers running, remove all containers first:
    docker rm $(docker container ls -aq);
    
    # Delete orphan (dangling) images only:
    docker rmi $(docker image ls -aq -f "dangling=true");
    
    # Then Rebuild (be sure to have the heroku cli installed in your machine)
    REBUILD=on ./scripts/serve-local.sh
    

• pipenv run ./joplin/manage.py shell_plus

---

Performance monitoring

We've set up Silk for performance monitoring of queries and memory usage. This is set up at the /performance endpoint. You'll need to login.

By default, Silk is set to run on every deployed environment besides production. If you want to use it locally, include MONITOR_PERFORMANCE=True in your .env file.

Load testing

We set up Locust to use for load testing. You can use it by running pipenv run locust. Be careful about where and how and when you do load testing. Be smart, don't crash staging or production (though you shouldn't be able to anyways since they have been load-tested).

The testing behavior is set up locustfile.py and can be expanded to do a wider variety of testing behaviors.

Debugging

Use the Django Debug Toolbar!

The django debug toolbar is super handy for debugging and performance testing. Simply run with the flag DEBUG_TOOLBAR=on, such as DEBUG_TOOLBAR=on ./scripts/serve-local.sh

With Pycharm

1. Run sh scripts/undockered.sh to initialize an undockered Joplin instance. This will run your initial data migration and seeding for you. It will also spin up joplin_db and joplin_assets containers. These are steps that our Pycharm debugging script can't do on its own.
2. Shut down ^C your undockered Joplin runserver. The joplin_db and joplin_assets containers should still be running.
3. Open Pycharm.
4. Open your 'Undockered Joplin' Run Configuration Run > Debug 'Undocked Joplin'. This run configuration should be git committed in your .idea/ folder. It will run a Joplin runserver command with the benefit of Pycharm's debugger.

Without Docker or Pycharm

1. Run sh scripts/undockered.sh to initialize an undockered Joplin instance.
2. Add import pdb; pdb.set_trace() to set a breakpoint & enjoy!

With Docker

1. Running pdb breakpoints inside the container is possible, but prone to complications. YMMV

---

Design

icons

To get a full set of icons that Wagtail has available you'll need to upload Wagtail's icomoon icon definitions to the icomoon web app. Make sure you're uploading the icon definitions for the version of wagtail we're using.

Adding Scripts/Styles

We're using webpack to bundle styles and scripts, and webpack_loader to include them in our templates. To create a new bundle it should be defined as an entry in webpack.build.js and webpack.dev.js, then included in a template using {% load render_bundle from webpack_loader %} and {% render_bundle 'YOUR_BUNDLE_NAME_HERE' %}.

---

Related Repos

There are a couple notable dependencies used by this project:

Forked projects

There are a couple existing projects that we use for Joplin and needed to fork to add functionality for our purposes. In these cases these forks are hopefully temporary, as we'd like to be able to contribute back to the main projects.

https://github.com/cityofaustin/django-countable-field

Provides support for multiple Django field types. It used used to add a character count widget to designated fields.

https://github.com/cityofaustin/wagtail-modeltranslation

We use wagtail-modeltranslation to handle translated fields, this fork corrects a single line of code to make it compatible with current and future versions of wagtail.

Publisher

https://github.com/cityofaustin/publisher

Microservice that handles publishing versions of Janis based on branches of Joplin.

coa-joplin-clean-up

https://github.com/cityofaustin/coa-joplin-clean-up

A webhook used to delete heroku PR builds after a PR has been closed or merged. If you want your PR branch to stay up even after being merged/closed, you can add the environment variable "DELETION_PROTECTION"=1 to your build.

Misc

Static File Uploads

The master branch (staging app, joplin-staging.herokuapp.com) and production branch (production app, joplin.herokuapp.com) upload static files to an S3 bucket (both share the same bucket), but have separate databases.

Syncing prod data to staging

1. heroku pg:copy joplin::DATABASE_URL DATABASE_URL -a joplin-staging

Changelog