Skip to main content

​
Implemented controls

  • Deterministic transfer identities and database uniqueness
  • BSC confirmation depth and reorg rewind
  • Contract-specific BSC checkpoints
  • Adaptive ranges with bounded retries
  • PostgreSQL advisory locking and in-process coordination
  • Durable synchronization runs and unresolved failure records
  • Partial-run reporting rather than all-or-nothing visibility
  • Liveness and database/configuration readiness endpoints
  • Graceful SIGINT/SIGTERM listener and pool shutdown

​
Required telemetry

Production metrics should cover safe-tip and cursor lag, run duration and state, processed ranges/pages, insert/update counts, adaptive reductions, RPC latency/rate limits, failure backlog and age, repeated replay, database pool saturation, stale readiness, and dashboard query latency. Logs should include a run ID, chain, source, bounded range/page, attempt, and classified error. Never include RPC credentials, session/OTP tokens, private keys, signed payloads, or full compliance evidence.

​
Scaling and recovery

Move scheduled indexing out of the API process and partition workers by chain/source under durable leases. Batch writes only after measuring correctness and query plans. Back up PostgreSQL with encryption and point-in-time recovery, define RPO/RTO, and test restoration. Derived chain data can be replayed from verified checkpoints, but an RPC provider is not a backup strategy.