Operations

Metrics endpoint

Vektor exposes lightweight, in-process operational metrics at /metrics in the Prometheus text exposition format, ready to scrape by Prometheus/Grafana or read by hand while debugging.

Browse documentation

Getting started

Reference

Operations

Metrics endpoint

Changelog

Latest updates

The /metrics endpoint

Every running server exposes metrics at /metrics. The endpoint lives outside the API router, so it is reachable even in API-only deployments and does not require authentication. Point your browser or curl at it:

Terminal

curl http://localhost:8080/metrics

The response uses the Prometheus text/plain; version=0.0.4 content type. Each metric is emitted with its # HELP and # TYPE lines followed by the current value:

/metrics

# HELP vektor_process_uptime_seconds Process uptime in seconds.
# TYPE vektor_process_uptime_seconds gauge
vektor_process_uptime_seconds 3821.44

# HELP vektor_http_requests_total Total number of HTTP requests handled since start.
# TYPE vektor_http_requests_total counter
vektor_http_requests_total 15204

# HELP vektor_websocket_active_connections Currently open realtime WebSocket connections.
# TYPE vektor_websocket_active_connections gauge
vektor_websocket_active_connections 7

Exposed metrics

The snapshot is collected on demand each time the endpoint is scraped. The following series are exposed:

Metric	Type	Description
`vektor_process_uptime_seconds`	gauge	Process uptime in seconds.
`vektor_http_requests_total`	counter	Total HTTP requests handled since start.
`vektor_http_requests_per_second`	gauge	Average HTTP requests per second over the last 60s.
`vektor_websocket_active_connections`	gauge	Currently open realtime WebSocket connections.
`vektor_spaces_total`	gauge	Number of active spaces.
`vektor_users_total`	gauge	Number of registered users.
`vektor_process_cpu_user_seconds_total`	counter	Total user CPU time consumed, in seconds.
`vektor_process_cpu_system_seconds_total`	counter	Total system CPU time consumed, in seconds.
`vektor_process_cpu_seconds_total`	counter	Total CPU time (user + system), in seconds.
`vektor_process_cpu_utilization_ratio`	gauge	Rolling 60s CPU usage as a fraction of one core (1.0 = one core fully busy).
`vektor_event_loop_delay_seconds`	gauge	Most recent event loop delay sample, in seconds.
`vektor_event_loop_delay_max_seconds`	gauge	Peak event loop delay over the last 60s. Sustained multi-second values mean the loop was blocked, stalling all clients.
`vektor_memory_rss_bytes`	gauge	Resident set size (total memory allocated for the process).
`vektor_memory_heap_total_bytes`	gauge	Total size of the V8 heap.
`vektor_memory_heap_used_bytes`	gauge	Used size of the V8 heap.
`vektor_memory_external_bytes`	gauge	Memory used by C++ objects bound to JavaScript objects.
`vektor_memory_array_buffers_bytes`	gauge	Memory allocated for ArrayBuffers and SharedArrayBuffers.

Prometheus scrape config

Add a scrape job pointing at your server to collect these metrics:

prometheus.yml

scrape_configs:
  - job_name: vektor
    metrics_path: /metrics
    static_configs:
      - targets: ["vektor.example.com:8080"]

From there, the series are ready to graph in Grafana or alert on. For example, alerting when vektor_event_loop_delay_max_seconds stays above a threshold, or tracking vektor_spaces_total and vektor_users_total growth over time.