Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/tines/rails-pg-adapter

Rails Postgres ActiveRecord patches for common production workloads
https://github.com/tines/rails-pg-adapter

activerecord availability monkeypatch postgresql rails

Last synced: 3 months ago
JSON representation

Rails Postgres ActiveRecord patches for common production workloads

Awesome Lists containing this project

README 

        
# RailsPgAdapter



This project allows you to monkey patch `ActiveRecord` (PostgreSQL) and auto-heal applications in production when PostgreSQL database fails over or when a cached column (in `ActiveRecord` schema cache) is removed from the database from a migration in another process.



## Installation



Install the gem and add to the application's Gemfile by executing:



    $ bundle add rails-pg-adapter



If bundler is not being used to manage dependencies, install the gem by executing:



    $ gem install rails-pg-adapter



## Usage



### Auto healing connections when PostgreSQL database fails over



```ruby

# config/initializers/rails_pg_adapter.rb



RailsPgAdapter.configure do |c|

  c.add_failover_patch = true

end

```



This will add the monkey patch which resets the `ActiveRecord` connections in the connection pool when the database fails over. The patch will reset the connection and re-raise the error each time it detects that an exception related to a database failover is detected.



### Retrying queries



When the database is failing you can retry queries that are not in a transaction. The gem will perform a back off retry in establishing the connection.

Once the back off is reached and no connection is found, it will bubble up the exception. Otherwise, the query will be retried with a new connection.



It is an opt-in functionality. You can supply your own back off figures for retries (in seconds) as following:



```ruby

# config/initializers/rails_pg_adapter.rb



RailsPgAdapter.configure do |c|

  c.add_failover_patch = true

  c.reconnect_with_backoff = [0.5, 1, 2, 4, 8, 16] # seconds

...

end



```



### Refresh model column information on the fly after an existing column is dropped



```ruby

# config/initializers/rails_pg_adapter.rb



RailsPgAdapter.configure do |c|

  c.add_reset_column_information_patch = true

end

```



This will clear the `ActiveRecord` schema cache and reset the `ActiveRecord` column information memoized on the model. The patch will reset the relevant information and re-raise the error each time it detects that an exception related to a dropped column is raised.



## How does it work



During a database failover in production, the `ActiveRecord` connection pool can become exhausted as queries are made against the database during the failover process. This can leave the `ActiveRecord` connection pools with stale or bad connections, even after the database has successfully recovered. Recovering from this issue usually requires a rolling restart of the application processes or containers.



`RailsPgAdapter` addresses this problem by resetting the connection pool and re-raises the original exception from an `ActiveRecord` monkey patch. This allows the application to auto-heal from stale connections on its own (after database recovery) when performing queries for a new request, without requiring manual intervention.



Another issue with `ActiveRecord` queries is `PG::UndefinedColumn`, which occurs when an `ActiveRecord` model includes a `SELECT` query with the name of a column that has been dropped from a Rails migration. This can happen even if the column isn't being referenced anywhere in the code. It occurs when a model is using `ignored_columns`, which prompts `ActiveRecord` to perform a dedicated lookup of the allowed columns in a select, such as `SELECT "users".name, "users".template_id...."`, instead of `SELECT "users".*`. When a column like `template_id` is dropped, PostgreSQL throws an undefined column error, which is bubbled up by `ActiveRecord` into `PG::UndefinedColumn`. Recovering from this issue also usually requires a rolling restart of the application processes or containers.



`RailsPgAdapter` solves this second issue by resetting the `ActiveRecord` schema cache and memoized model column information when it detects a `PG::UndefinedColumn` raised from a monkey patch. Resetting the column information forces `ActiveRecord` to refresh its schema cache by loading the table information from the database and no longer reference the dropped column for new queries, without requiring manual intervention.



## Development



- Install ruby 3.0



```

\curl -sSL https://get.rvm.io | bash



rvm install 3.0.0



rvm use 3.0.0

```



- `docker compose up -d` - to spin up postgres locally

- `bundle exec rspec` to run the tests.

- You can also run `bin/console` for an interactive prompt.



To install this gem onto your local machine, run `bundle exec rake install`.



## Contributing



Bug reports and pull requests are welcome on GitHub at https://github.com/tines/rails-pg-adapter. This project is intended to be a safe, welcoming space for collaboration, and contributors are expected to adhere to the [code of conduct](https://github.com/tines/rails-pg-adapter/blob/main/CODE_OF_CONDUCT.md).



## License



The gem is available as open source under the terms of the [MIT License](https://opensource.org/licenses/MIT).



## Code of Conduct



Everyone interacting in the `RailsPgAdapter` project's codebases, issue trackers, chat rooms and mailing lists is expected to follow the [code of conduct](https://github.com/tines/rails-pg-adapter/blob/main/CODE_OF_CONDUCT.md).



## Releasing a new version



- Bump version in `version.rb`

- Update `CHANGELOG.md`

- Push the changes to `main`

- Run the release script with the new version `./bin/release.sh 0.2.0`

  - Note: It will ask for MFA.

- Create a new release - https://github.com/tines/rails-pg-adapter/releases/new