Appearance
Redundancy and failover
Run at least two nodes with the same VRF key:
| Primary | Backup | |
|---|---|---|
| VRF key | X | same X |
| Transaction account | A | own account B |
extra_confirmations | 0 | 2 |
| Host and RPC providers | 1 | different |
The coordinator accepts only the first valid proof for a request, so double fulfilment is impossible.
- Primary healthy: the backup sees each request already fulfilled and sends nothing.
- Primary down: the backup answers 2 blocks later, about 20 s extra on PulseChain.
Drill results (PulseChain testnet, October 2026): with both nodes up, the backup sent zero transactions. With the primary stopped, the backup fulfilled the request 82 s after it was made.
Set fallback_rpc_urls on every node. Each request goes to the best-ranked provider and moves on only if that one errors. Providers are never queried in parallel, so a lagging provider cannot hide events.
Full procedure: verity-node/docs/runbooks/failover.md.