Operators / Monitoring

watch the signals that matter.

Monitoring should make operator action obvious before degraded infrastructure becomes a network or user-facing incident.

core signals

chain synchronization and block progress
peer count and connectivity quality
CPU, memory, disk usage and disk latency
RPC latency, error rates and saturation
indexer lag, failed jobs and data freshness
validator participation and missed duties where applicable

alert design

Alert on actionable conditions rather than raw noise. Thresholds should include duration and severity so brief fluctuations do not trigger the same response as sustained failure.

logs & retention

Centralize logs with timestamps, host identity and service version. Protect logs from accidental deletion while avoiding retention policies that exhaust production storage.

incident context

Record deploys, upgrades and configuration changes alongside telemetry so operators can correlate failures with recent changes and recover more quickly.