Prodder

A tool to maintain and load your Rails application's database structure, seed table contents, permissions and database settings based on its migration history and the current state in production databases.

In short: db:reset db:migrate

1. Your project maintains db/structure.sql, db/seeds.sql, and optional db/quality_checks.sql and db/permissions.sql files as it sees fit (ie, by using prodder as a script to dump production and push it to your git repository).
2. Make sure db/seeds.sql includes the schema_migrations table.
3. Only new migrations will be run against prod's structure using its seed table contents.
4. Once a migration has been deployed, it should result in db/structure.sql and db/quality_checks.sql files being modified, and any new seed data being added to db/seeds.sql -- including the new entry in schema_migrations.
5. That migration never needs to be run in development again. Feel free to rm.
6. Any application related permission changes will result in db/permissions.sql being modified.

prodder can be included as a railtie in your application to automatically replace many of Rails' db:* tasks. The only prerequisites to its usage are the existence of db/structure.sql, db/seeds.sql with at least the schema_migrations table contents included. Optional db/quality_checks.sql and db/permissions.sql will be loaded after seeding, which can be helpful if you wish to seed the database prior to enforcing foreign key constraints and if you want to develop in an environment with the same permissions setup as production.

In your Gemfile:

gem 'prodder', require: 'prodder/railtie'

It doesn't really matter, but for sanity's sake, you should set your schema_format to :sql:

# config/application.rb
module Whatever
  class Application
    config.active_record.schema_format = :sql
  end
end

If you want to work with permissions setup like production:

# config/database.yml

username: user_who_will_run_your_app (# for example identity__web)
migration_user: migration_overlay_user_who_is_usually_db_owner (# for example identity__owner)
superuser: godmode (# for example postgres)

Note that the migration_user and superuser must be created before you run db:reset db:migrate, just as you would have to when you have to create the username: user even with the standard setup. Note that this is only ever recommended for development environments. Do NOT mess with production overlays to have 3 different users in production, that's the reason for having overlays in the first place.You can also easily control this by making the gem a group :development only dependency:

#!/usr/bin/env bash

set -eu

createuser --superuser --createrole --createdb godmode || true
createuser --createrole migration_overlay_user_who_is_usually_db_owner || true

bundle install
bundle exec rake db:reset db:migrate

Things that really matter:

1. rake db:reset recreates your database, loading db/structure.sql, db/seeds.sql, db/quality_checks.sql and permissions.sql.
2. rake db:migrate runs migrations, as before, but runs after the initial seeds were created. Those initial seeds should have included your production app's schema_migrations table contents. This means only those migrations that have not yet run in production will need to be run locally.
3. If you configured to have 3 users in your #config/database.yml file and have a permissions.sql file present, all your db:* commands will be run in the context of the user it makes the most sense to run as, mimicking our production environment. For instance(s), to reset the database (god forbid we do this in production), it will run as superuser, to run a migration, as the migration_user and your application will connect to the database as username. Thus it achieves the overlays of a DBA, migration and production application.
4. Having 3 users configured and to achieve the effects of step 3, you must have a permissions.sql. However, you do not need to have 3 users configured to restore permissions (load the permissions.sql file). This being said, it does not make sense to restore permissions in your environment if you're just going to run everything as a single, most likely superuser.

This will remove the db:* tasks:

• db:_dump: an internal task used by rails to dump the schema after migrations. Obsolete.
• db:drop:*
• db:create:*
• db:migrate
• db:migrate:reset
• db:migrate:up
• db:migrate:down
• db:fixtures:.*
• db:abort_if_pending_migrations
• db:purge:*
• db:charset
• db:collation
• db:rollback
• db:version
• db:forward
• db:reset
• db:schema:*
• db:seed
• db:setup
• db:structure:*
• db:test:*
• test:prepare: Rails 4.1 added this task to auto-maintain the test DB schema.

And reimplement only the following:

• db:structure:load: Load the contents of db/structure.sql into the database of your current environment.
• db:seed: Load db/seeds.sql into the database of your current environment.
• db:quality_check: Load db/quality_checks.sql into the database of your current environment, if present.
• db:reset: db:drop db:setup
• db:settings: Load the contents of db/settings.sql into the database of your current environment.
• db:setup: db:create db:structure:load db:seed db:quality_check db:settings
• db:test:prepare: RAILS_ENV=test db:reset db:migrate
• db:test:clone_structure: RAILS_ENV=test db:reset db:migrate
• test:prepare: db:test:prepare
• db:drop: Drop database as superuser
• db:create: Create database as superuser and transfer ownership to migration_user
• db:migrate:*, db:rollback Run migrations up/down as migration_user
• db:purge:*, db:charset, db:collation, db:version, db:forward, db:rollback, db:abort_if_pending_migrations as appropriate users.

See lib/prodder/prodder.rake for more info.

This is likely to cause issues across Rails versions. No other choice really. It has been used in anger on Rails 3.2.x and Rails 4.1.x.

Confirmed working versions of Postgres:

• 9.1.11+
• 9.2.6+

prodder is configured using a simplistic YAML file. There is no logic performed to locate that file; the path to it must be provided each time prodder is invoked.

# Each top-level key is the name of the Rails project for which we are updating
# the structure and seed data. Let's pretend we're maintaining a online store.

store:
  structure_file: db/structure.sql
  seed_file: db/seeds.sql
  quality_check_file: db/quality_checks.sql
  permissions:
    file: db/permissions.sql
    included_users: service__owner, service__web, prodder__read_only
  git:
    origin: git@github.com:pd/store.git
    author: prodder auto-commit <pd+prodder@krh.me>
  db:
    name: store_production_db
    host: store-db.krh.me
    user: prodder_readonly
    password: super-secret
    tables:
      - schema_migrations
      - coupons
      - products
      - vendors
    exclude_tables:
      - production_only_table
    exclude_schemas:
      - production_only_replication_schema

If you would prefer to maintain the list of seed tables within your application itself, the db.tables key can be given the path to a YAML file from which to load the list instead:

# prodder-config.yml
store:
  # [snip]
  db:
    tables: config/seeds.yml

# store/config/seeds.yml:
- schema_migrations
- coupons
- products
- vendors

In some cases, such as foreign key dependencies and triggers, you may wish to defer loading constraints on your tables until after your seed data has been loaded. prodder treats the presence of a quality_check_file key in the configuration as an indication that it should split structure_file into those statements which create the base structure, and put the constraints into the quality_check_file.

We have had multiple cases in the past with deployments failing because some role cannot access something on prod. To fail early and catch these in development, it would be easier to just have these permissions loaded in development environments. However, note that to actually take advantage of these restored permissions, you must configure the 3 users as mentioned before in #config/database.yml.

store:
  structure_file: db/structure.sql       # CREATE TABLE ...
  quality_check_file: db/quality_checks.sql # ALTER TABLE ... ADD FOREIGN KEY ...

The -c option to specify the configuration file is always required. All options should be passed at the end of the command line.

# Clone all the repositories in prodder.yml, update the remotes for
# any that already existed but have new origins specified, and
# check out the branch specified in the config. The branch must be
# the name of an available remote branch, as the local branch will
# be forcibly reset to its current SHA with `git reset --hard`.
$ prodder init -c prodder.yml

# Dump the remote databases' structures and seed tables into the
# files specified by structure_file and seed_file.
$ prodder dump -c prodder.yml

# Lots of projects? Dump just one or two:
$ prodder dump store chordgen -c prodder.yml

# Commit the changes to each project's repository. Only repositories
# with actual changes will be committed.
$ prodder commit -c prodder.yml

# Push.
$ prodder push -c prodder.yml

• Log activity as it is performed.
• Support tracking a particular branch instead of master.
• Support specifying the options to pass to each pg_dump form.
• Select dumping only a subset of a seed table. (pg_dump won't do this ...)

• Kyle Hargraves
• Sri Rangarajan
• Emmanuel Sambo
• Cindy Wise
• Robert Nubel
• Josh Cheek

The gem is available as open source under the terms of the MIT License.