Case study · Software / Open source
Agrafa
Know the state.Catch the change.
A Go-based monitoring system that collects node and service health, evaluates state, applies alert rules and exposes a useful read model for dashboards.
- Agent
- Go / gopsutil
- Backend
- Go / PostgreSQL / REST
- Focus
- Nodes / services / alerts
01 / Overview
Monitoring is not the graph.
It is knowing what needs attention. Agrafa collects bounded operational signals, derives current state, detects meaningful changes and makes the result readable without a query language.
It is open-source project work, not a commercial observability product.
02 / Under the hood
Signal → state → action.
The backend turns raw telemetry into state, so the interface does not have to reconstruct meaning from individual data points.
- 01
Agent
Collects heartbeat, health and bounded CPU, memory and disk metrics from each node.
- 02
Ingestion
Receives agent payloads through focused REST endpoints and stores the observations.
- 03
State engines
Evaluate node and service state, including heartbeat expiry.
- 04
Rules + events
Turn state changes and threshold breaches into alerts and a useful event history.
03 / Current status
Enough system to prove the idea.
The core monitoring path works. The interesting decisions now are about where it should go next.
- Alerts
- Offline, unhealthy and threshold rules
- Case study
- Full write-up in progress