Prometheus

Storage Map v101 → current

Prometheus endpoint info for neurons.

Explore chain
Queried by: validatorssubnet ownersdevelopersanalytics

The Big Picture

Prometheus endpoints expose operational metrics from neurons - inference latency, request counts, error rates, resource usage, etc. This enables observability across the network. Validators can use this data to make better weight decisions, and operators can debug issues.

Why This Matters

Beyond just connecting to miners, you want to know how they're performing. Prometheus endpoints give you operational metrics - how fast, how reliable, how loaded is this neuron?

Example Scenario

Query Prometheus(netuid=1, uid=47) returns { ip: '1.2.3.4', port: 9090, ip_type: 4, version: 1, block: 7000000 }. Scrape http://1.2.3.4:9090/metrics for Prometheus-format metrics.

Common Questions

Is Prometheus required?
No, it's optional. Neurons can participate without exposing metrics. But providing metrics can build trust with validators who value transparency.
What metrics should neurons expose?
Common: inference_latency, request_count, error_rate, queue_depth. Subnet-specific metrics depend on the task. Follow Prometheus naming conventions.
Can validators use this for scoring?
Yes, sophisticated validators can scrape metrics to inform weight decisions - rewarding neurons with better performance characteristics.

Use Cases

  • Build network-wide monitoring dashboards
  • Scrape neuron performance metrics for analysis
  • Debug neuron health and performance issues
  • Create alerting systems based on neuron metrics
  • Research network behavior through metrics aggregation

Purpose & Usage

Purpose

Store metrics endpoint location for monitoring and observability.

Common Query Patterns

  • Query by netuid-uid for a specific neuron's metrics endpoint
  • Iterate to discover all available metrics endpoints
  • Build monitoring systems that scrape neuron metrics

Query Keys

#NameTypeDescription
1
key1
u16 key1 (u16)
2
key2
AccountId key2 (AccountId) (hex -> SS58)

Stored Value

PrometheusInfo

value (PrometheusInfo)

Relationships

Code Examples

import { ApiPromise, WsProvider } from "@polkadot/api";
import { stringCamelCase } from "@polkadot/util";

const provider = new WsProvider("wss://entrypoint-finney.opentensor.ai:443");
const api = await ApiPromise.create({ provider });

// Query Prometheus storage
const key1 = 0;
const key2 = "5GrwvaEF5zXb26Fz9rcQpDWS57CtERHpNehXCPcNoHGKutQY";

const result = await api.query
  [stringCamelCase("SubtensorModule")]
  [stringCamelCase("Prometheus")](
  key1,
  key2
);

console.log("Prometheus:", result.toHuman());

Runtime Info

View Source
Pallet
SubtensorModule
Storage Kind
Map
First Version
v101
Current Version
v393
Back to SubtensorModule Storage