Prometheus Skill

Prometheus metrics querying and monitoring operations.

Overview

This skill provides comprehensive access to the Prometheus HTTP API for querying metrics, managing alerts, and monitoring system status.

Requirements

•Prometheus server accessible via HTTP
•Optional: Basic authentication credentials if enabled

Configuration

bash

export SKILL_PROMETHEUS_URL=http://localhost:9090

Tools (15)

Query Tools

•query - Execute instant PromQL query at a single point in time
•query-range - Execute PromQL query over a time range with step interval
•series - Find time series matching label selectors
•labels - Get all label names in the database
•label-values - Get all values for a specific label

Target & Rule Tools

•targets - Get current scrape targets and their health status
•rules - Get alerting and recording rules
•alerts - Get currently firing alerts
•metadata - Get metric metadata (type, help, unit)

Status Tools

•status-config - Get current Prometheus configuration
•status-flags - Get configured flag values
•status-runtimeinfo - Get runtime information (memory, goroutines, etc.)
•status-buildinfo - Get build information (version, revision, etc.)
•status-tsdb - Get TSDB storage statistics
•check-health - Check if Prometheus is healthy and ready

Example Usage

Query Current Metrics

code

Tool: query
Args: { "query": "up" }
Result: Shows which targets are up (1) or down (0)

Query CPU Usage Over Time

code

Tool: query-range
Args: {
  "query": "rate(node_cpu_seconds_total{mode=\"user\"}[5m])",
  "start": "2024-01-01T00:00:00Z",
  "end": "2024-01-01T01:00:00Z",
  "step": "1m"
}
Result: CPU usage rate over the specified time range

Find Series by Label

code

Tool: series
Args: { "match": "{job=\"prometheus\"}" }
Result: All time series for the prometheus job

Get Label Values

code

Tool: label-values
Args: { "label": "job" }
Result: ["prometheus", "node-exporter", "alertmanager", ...]

Check Firing Alerts

code

Tool: alerts
Args: {}
Result: List of currently firing alerts with labels and annotations

Get Target Status

code

Tool: targets
Args: { "state": "active" }
Result: All active scrape targets with last scrape time and health

PromQL Query Examples

Basic Queries

•up - Target availability
•node_memory_MemTotal_bytes - Total memory
•process_cpu_seconds_total - Process CPU time

Rate Calculations

•rate(http_requests_total[5m]) - Request rate over 5 minutes
•irate(http_requests_total[5m]) - Instantaneous request rate

Aggregations

•sum(rate(http_requests_total[5m])) by (handler) - Requests by handler
•avg(node_cpu_seconds_total) by (instance) - Average CPU by instance

Filtering

•http_requests_total{status="500"} - Only 500 errors
•node_disk_io_time_seconds_total{device=~"sd.*"} - Disk devices matching pattern

Response Format

All query results follow the Prometheus API response format:

json

{
  "resultType": "vector|matrix|scalar|string",
  "result": [
    {
      "metric": { "label": "value" },
      "value": [timestamp, "value"]
    }
  ]
}

Error Handling

•Connection errors return descriptive messages
•Invalid PromQL queries return Prometheus error messages
•Timeout errors indicate slow queries (consider adjusting timeout parameter)

Security Notes

•No authentication is performed by default
•Configure Prometheus with authentication if exposed publicly
•Queries can be resource-intensive; use appropriate time ranges and step intervals