Handbook to migrate your Postgres from Heroku to Kamal

Today I’m going to show you how I managed to migrate from Heroku to Kamal, with a focus on Postgres. The schema below describes our goal for today :

We want :

1. A Preprod application, on its own host and its own Postgres instance.
2. A Production application, with its own Postgres as well. Production also needs another feature : save DB backups on an S3.

We will use my previous article as a base for today. You can find every information here.

Table of Contents

 1. Add Postgres to Production

       1.1. Edit deploy.yml
       1.2. Edit secrets
       1.3. Edit database.yml

 2. Setup Postgres for Preprod

       2.1. Edit deploy.preprod.yml
       2.2. Edit secrets
       2.3. Edit database.yml

 3. Migrate the database

       3.1 Download the backup
       3.2. Upload the backup to S3
       3.3. Run the migration script

 4. Conclusion

Add Postgres to Production

We already have Production and Preprod applications working perfectly without any additional services. Let’s add and configure our first accessories : Postgres and Backups !

Edit deploy.yml

# config/deploy.yml
[...]
service: my-app

env:
  clear:
    POSTGRES_USER: postgres
    POSTGRES_DB: my_app_production
    POSTGRES_HOST: my-app-db
  secret:
    - RAILS_MASTER_KEY
    - POSTGRES_PASSWORD

accessories:
  db:
    image: postgres:16.4
    host: 190.0.0.0
    port: 5432:5432
    env:
      clear:
        POSTGRES_USER: postgres
        POSTGRES_DB: my_app_production
      secret:
        - POSTGRES_PASSWORD
    directories:
      - data:/var/lib/postgresql/data

  db-backup:
    image: eeshugerman/postgres-backup-s3:16
    host: 190.0.0.0
    env:
      clear:
        SCHEDULE: '@daily' # I want a backup every day for the last 30 days
        BACKUP_KEEP_DAYS: 30
        S3_REGION: eu-west-3
        S3_PREFIX: backups
        S3_BUCKET: my-app-backups
        POSTGRES_HOST: my-app-db
        POSTGRES_DATABASE: my_app_production
        POSTGRES_USER: postgres
      secret:
        - POSTGRES_PASSWORD
        - S3_ACCESS_KEY_ID
        - S3_SECRET_ACCESS_KEY

As my service name is my-app, Kamal will automatically prefix accessories with it. This is why I can freely call my-app-db. Kamal understands that you’re talking about the accessory db linked with the service my-app. This mechanism is very powerful!

For the database backups, I’ll use the marvelous image that does all the work for you : https://github.com/eeshugerman/postgres-backup-s3

Basically, you just need to set up an S3 with the correct permissions, and every day you’ll get your backups there. postgres-backup-s3 also provide a useful feature that allows you to restore an old backup. We will use it later on !

Edit secrets

We now need to ensure that the new secrets are stored in the secrets file:

# .kamal/secrets
[...]
POSTGRES_PASSWORD=fakepassword
S3_ACCESS_KEY_ID=xxxxxxxxxxxxxxx
S3_SECRET_ACCESS_KEY=xxxxxxxxxxxxxxx

Edit database.yml

You also need to help Rails connect to your database by providing the same environment variables:

# config/database.yml
default: &default
  adapter: postgresql
  pool: <%= ENV.fetch("RAILS_MAX_THREADS") { 5 } %>
  timeout: 5000
  host: <%= ENV["POSTGRES_HOST"] %>
  username: <%= ENV["POSTGRES_USER"] %>
  password: <%= ENV["POSTGRES_PASSWORD"] %>

[...]

production:
  <<: *default
  database: <%= ENV.fetch("POSTGRES_DB") { "my_app_production" } %>

Once it’s done, you can run :

$ kamal setup

This will install the Postgres container in Production. You can confirm that everything will work by accessing the dbconsole using : kamal app exec -i 'bin/rails db'

Now that we have Postgres working in Production, let’s do the same for Preprod.

---

Setup Postgres for Preprod

Well, the steps are exactly the same. We will just provide different env variables. Let’s do it!

Edit deploy.preprod.yml

# config/deploy.preprod.yml
[...]
service: my-app-preprod
[...]
env:
  clear:
    POSTGRES_USER: postgres
    POSTGRES_DB: my_app_preprod
    POSTGRES_HOST: my-app-preprod-db
  secret:
    - RAILS_MASTER_KEY
    - POSTGRES_PASSWORD

accessories:
  db:
    image: postgres:16.4
    host: 190.0.0.1
    port: 5432:5432
    env:
      clear:
        POSTGRES_USER: postgres
        POSTGRES_DB: my_app_preprod
      secret:
        - POSTGRES_PASSWORD
    directories:
      - data:/var/lib/postgresql/data

The only changes from Production are :

— service property.
— POSTGRES_DB and POSTGRES_HOST env variables.
— Host domains for accessories

Edit secrets

We still need to update .kamal/secrets.preprod to provide a POSTGRES_PASSWORD value :

# .kamal/secrets.preprod
[...]
POSTGRES_PASSWORD=fakepassword

Edit database.yml

We need to add the preprod environment and confirm that it takes the right environment variable.

# config/database.yml
[...]
preprod:
  <<: *default
  database:  <%= ENV.fetch("POSTGRES_DB") { "ride_back_preprod" } %>

And, we’re good to run :

$ kamal setup -d preprod

We now have Preprod and Prod applications working with their own database 🎉

Now that everything is set up, let’s say adios to Heroku and migrate our database once and for all!

---

Migrate the database

The first thing you want to do is cut access to your application to avoid losing any data.

To do so, you can enable the “Maintenance Mode” in the settings of your Heroku application.

Once it’s done, we can get back to work !

Download the backup

You’ll need to find your Postgres public URL to download a backup locally. To do so, go on your application Dashboard and :

“Resources” > “Heroku Postgres” > “Credentials” > “Default” > “URI”

The URI will normally look like this : postgres://{random-letters-and-numbers}ec2[...].eu-west-1.compute.amazonaws.com:5432/[...]
Once you’ve found it, let’s download the database locally using pg_dump:

pg_dump --format=c --no-acl --no-owner {paste URI here}  > database.dump

You should now see a database.dump file in your directory.

Upload the backup to S3

Now, you need to upload your file to the S3 bucket you use to store your backups. Use the exact same directory the backups will go.

Run the migration script

Once your file is uploaded, you can now run this script:

kamal accessory exec db-backup -i "sh" 
source ./env.sh
s3_uri_base="s3://${S3_BUCKET}/${S3_PREFIX}"
aws $aws_args s3 cp "${s3_uri_base}/database.dump" "database.dump"
conn_opts="-h $POSTGRES_HOST -p $POSTGRES_PORT -U $POSTGRES_USER -d $POSTGRES_DATABASE"
pg_restore $conn_opts --clean --if-exists --no-acl --no-owner database.dump
rm -rf database.dump

This script will:

1. Access to the database backup container using sh
2. Download your database.dump file using aws cli
3. Restore the database based on database.dump
4. Remove database.dump to free space.

That’s it! Your database is now completely self-hosted. Congratulations!

Conclusion

Thanks for reading so far !

If you need help with your Kamal configuration, do not hesitate to comment out or reach me by DM 👋

Comments

[Thomas Frank]

I appreciate the emphasis on data security and disaster recovery by utilizing the postgres-backup-s3 image for automated backups. This demonstrates a proactive approach to managing the database and ensures that data is protected in case of unforeseen Incredibox Colorbox Mustard events.