We understand that the reliability of our High-Availability CDN is critical for your production GraphQL Gateway. Even a brief outage can have a significant impact.
To further enhance the resilience of our systems, we’ve introduced two major improvements to mitigate potential outages of critical components for the Gateway.
CDN Mirror for High Availability
We now operate a second CDN mirror, cdn-mirror.graphql-hive.com, built on AWS CloudFront. This
mirror serves as a complete replica of our primary Cloudflare-based CDN. In the event of an outage
with our main CDN, you can seamlessly switch to the mirror, ensuring your schemas and artifacts
remain available.
Our official SDKs have been updated to automatically handle this fallback when configured with both endpoints.
Circuit Breaker for Usage Reporting and CDN requests
Failures are inevitable. To gracefully handle transient issues with our usage reporting service, we’ve implemented a Circuit Breaker pattern in our Hive SDKs and the Hive Gateway.
If the client detects a series of failed requests to any origin, it will temporarily stop sending new request to that origin. This prevents your gateway or service from being overwhelmed with failing requests and allows it to run stable even if an outage on the CDN or usage reporting occurs.
Conclusion
These improvements are now available for the Hive JS SDK and JavaScript Hive Gateway. We will soon also ship these improvements for the Hive Rust SDK, Hive Router and the Apollo Router Hive Plugin.
Please upgrade to the following versions, and adjust your configuration to use the dual CDN.
- Hive Gateway
v2.1.22 @graphql-hive/yoga@0.46.0@graphql-hive/core@0.18.0@graphql-hive/apollo@0.45.0
These updates are part of our ongoing commitment to providing a highly available and reliable schema registry for your GraphQL APIs.
Further reading: