Failsafe
Routing decides which Endpoint to try. Failsafe decides how many times to try, when an operation can move to another Endpoint, and when to stop sending traffic to an Endpoint at all. It covers attempt bounds, operation-aware failover, the Circuit, health observations, and response validation.
In this chapter
| Page | Covers |
|---|---|
| Retries | max_attempts and the failure and operation semantics that move a request to the next candidate |
| Circuit | when the Gateway stops sending an Endpoint traffic, and how it recovers |
| Health observations | where observational latency and health status come from |
| Response validation | the structural checks applied before a response is returned |
Runtime model
- Attempts run sequentially against distinct Endpoints; the client owns whole-request deadlines, backoff, and jitter.
- Each upstream result is handled independently. The Gateway forwards
result: nulland valid upstream JSON-RPC application responses unchanged. - Circuit admission uses connection and response observations. Applications apply chain-specific tip-lag, consensus, checksum, and confirmation policies.
- Route configuration supplies candidates and the attempt bound. The Gateway classifies failure certainty and operation replay behavior internally; client retry logic adds its own bounded request-level policy around the Gateway call.
Related
Last updated on