Solana Exporter
A public guide to the metrics that matter for Solana and Agave validator environments.
The Solana exporter should represent validator health, voting, epoch timing, and performance, not just whether the process is running. It complements host metrics; it does not replace them.
For validator version compliance and epoch-oriented maintenance visibility, see Solana-EpochRoll.
Use the Asymmetric Research Solana exporter for RPC-derived metrics. For Kubernetes deployments, supply its RPC endpoint, validator identities, and collector configuration through your own reviewed deployment configuration.
Scrape targets
Use separate jobs or clear labels for:
- Validator/RPC JSON-RPC-derived metrics.
- Node exporter metrics for CPU, memory, disk, filesystem, and network interfaces.
- Client-native Prometheus endpoints where a validator client exposes one.
- Blackbox probes for externally reachable RPC and WebSocket endpoints.
Avoid exposing the exporter itself to the public internet. Prometheus should scrape it over an approved network path, with a defined authentication policy where appropriate.
An exporter Helm chart commonly runs one replica, passes its RPC URL and node/vote key configuration as arguments, exposes a metrics container port through a ClusterIP Service, and makes Prometheus discovery opt-in. Enable and label the ServiceMonitor deliberately in an environment overlay; otherwise installing a chart alone does not create Prometheus discovery.
Minimum dashboard panels
| Panel | Why it exists |
|---|---|
| Local versus reference slot | Detects falling behind before the node becomes obviously unavailable. |
| Vote-account status and delinquency | Separates a reachable node from a functioning validator. |
| Epoch position and leader schedule context | Helps choose maintenance windows. |
| Block production and missed slots | Shows validator quality during assigned work. |
| Ledger/accounts/snapshot disk use | Makes capacity and failed snapshot risks visible. |
| UDP/network errors and host pressure | Connects protocol symptoms to machine-level causes. |
| Exporter scrape errors and last successful collection | Makes monitoring failure visible. |
Alert ownership
Each alert needs a destination, severity, runbook link, and maintenance-suppression rule. An alert that only says “Solana down” is not actionable; include the endpoint, cluster, observed/reference slot, and the first diagnostic command.
The full metric-to-RPC map, derived metrics, and alert design is maintained in the Solana / Agave section.