Add database connection resilience to the indexer

[rustydb]

Priority: P1

Source

Senior Architect Code Review (2026-04-03)

Problem

The Rust indexer uses a single tokio_postgres client connection with no reconnection logic or connection pooling. If the database connection drops (network blip, Cloud SQL restart, maintenance window), the indexer crashes and requires a full container restart.

The current code spawns the connection task and logs errors but does not attempt recovery:

tokio::spawn(async move {
    if let Err(error) = connection.await {
        eprintln!("postgres connection error: {error}");
    }
});

Affected Files

• apps/indexer/src/main.rs (Database struct)

Recommendation

Replace the bare tokio_postgres client with a connection pool such as deadpool-postgres or bb8-postgres. These handle:

• Automatic reconnection on connection loss
• Connection health checking
• Configurable pool size for concurrent package pollers
• Graceful degradation during database outages

Alternatively, wrap the existing client in a reconnection loop that detects connection failures and re-establishes the connection before retrying.

Acceptance Criteria

• Indexer survives a temporary database outage without crashing
• Connection is re-established automatically after a drop
• Concurrent package pollers share a connection pool
• Connection pool size is configurable via environment variable
• Behavior is tested (at minimum, a reconnection scenario)