Checklist
Ethereum RPC production checklist
What to verify before an Ethereum RPC endpoint carries production traffic: sync, pruning, failover, monitoring and load balancing.
Nodes
- Execution and consensus clients are on supported versions, with a written upgrade path.
- Each node's role is explicit: archive, full or minimal, and the routing reflects it.
- Pruning and history settings are deliberate and documented, including Erigon prune distance.
- Disk has headroom for growth and for a resync, on storage that can sustain the write load.
Correctness
- A node that lags the chain head is removed from rotation automatically.
- Block height is compared across nodes, not only checked against one node's own view.
- Re-orgs are handled: responses are not served from an abandoned fork.
- Stateful methods such as filters and subscriptions keep affinity to one node.
Load balancing and failover
- Requests are routed by method and capability, not round-robin.
- Expensive calls such as trace and log queries cannot starve ordinary traffic.
- Rate limits exist per client, and a failed node triggers failover that has been tested.
Monitoring
- Alerts fire on sync lag, peer count and error rate, not only on host health.
- Per-method latency and error metrics are available.
- There is a runbook for a stuck sync and for a resync from scratch.
Want someone to run through this with you?
The Ethereum node and RPC health check does exactly that: 840 – 1,680 EUR, Scoped on the call.