activerecord-trilogy-failover

Automatic reconnection on MySQL read-only errors for ActiveRecord's Trilogy adapter.

When a MySQL server switches to read-only mode (e.g., Aurora failover, ProxySQL routing change, RDS Multi-AZ switchover), existing connections receive ER_OPTION_PREVENTS_STATEMENT (1290) on write attempts. This gem translates that error into ActiveRecord::ConnectionFailed, enabling Rails' built-in retry mechanism to transparently reconnect to the new writer.

The gem prepends a thin patch to TrilogyAdapter#translate_exception:

Write attempt → 1290 error → ConnectionFailed → Rails retry → reconnect! → new writer

• SELECT (with allow_retry: true): Automatically retried and reconnected
• INSERT / UPDATE / DELETE: Error propagates to the caller (no double-execution risk)

This is the correct behavior — writes should not be silently retried.

Add to your Gemfile:

gem "activerecord-trilogy-failover", github: "riseshia/activerecord-trilogy-failover"

With Rails: No configuration needed. The Railtie automatically patches TrilogyAdapter on load.

Without Rails: Manually prepend the patch:

require "activerecord_trilogy_failover"

ActiveRecord::ConnectionAdapters::TrilogyAdapter.prepend(
  ActiveRecordTrilogyFailover::AdapterPatch
)

Rails' built-in retry settings work with this gem:

production:
  adapter: trilogy
  connection_retries: 2    # default: 1
  retry_deadline: 5        # seconds, default: none

• Ruby >= 3.2
• ActiveRecord >= 7.1
• Trilogy >= 2.0
• MySQL 8.0 (other versions are not tested)

bundle install
bundle exec rspec

Integration tests run against a real MySQL instance using Docker:

docker compose up -d            # start MySQL on port 3307
MYSQL_PORT=3307 bundle exec rspec spec/integration/
docker compose down -v          # cleanup

MIT