In our production setup, we use Apollo Gateway (with NestJS) to federate multiple GraphQL subgraphs. Our biggest challenge is safely updating the federated schema in a running Gateway instance without restarting the entire process. Polling also fails silently so i want reliable solution to update schema on gateway without restart.
What doesn’t work reliably:
- Polling-based schema reloading (
gateway.load()+ replacing schema in memory) leads to errors and crashes in NestJS-integrated environments. - Restarting the entire Gateway app just to reflect subgraph schema changes is not scalable or production-friendly, especially when uptime is critical.
What we’re asking for:
We are looking for an officially supported, safe method to re-initialize or update the schema in Apollo Gateway:
- Without full process restarts
- Without causing transient failures for live GraphQL requests
- Ideally triggered by polling, webhook, or registry change
Why it matters:
This capability is vital for CI/CD pipelines and dynamic microservice environments where subgraphs may evolve independently. Without this, deployments are brittle, and downtime is unavoidable.