Prometheus metrics¶
Waylight can expose its measurements at GET /metrics in the Prometheus
text format, so an existing Prometheus/Grafana setup can scrape them. The
endpoint is off by default.
Enable¶
# .env
METRICS_ENABLED=true
METRICS_TOKEN=<at least 32 characters, e.g. the output of: openssl rand -hex 32>
The endpoint requires Authorization: Bearer <METRICS_TOKEN> and is, like
every other route, limited to ALLOWED_CIDRS. It exposes infrastructure
details (service names, container names, disk sizes), so keep the token
secret and do not publish the port.
Scrape config¶
scrape_configs:
- job_name: waylight
scrape_interval: 60s
metrics_path: /metrics
authorization:
type: Bearer
credentials_file: /etc/prometheus/waylight.token
static_configs:
- targets: ["192.168.178.115:8090"]
A ready-made Grafana dashboard is in
grafana-dashboard.json (Dashboards → New →
Import, then pick your Prometheus data source).
Metrics¶
All names are stable; labels have a small, bounded set of values (names you configured yourself), never request IDs or paths.
| Metric | Type | Labels | Meaning |
|---|---|---|---|
waylight_build_info |
gauge | version, goversion |
Always 1 |
waylight_service_up |
gauge | service |
1 if the last health check succeeded |
waylight_service_response_seconds |
gauge | service |
Response time of the last check |
waylight_service_cert_expiry_timestamp_seconds |
gauge | service |
TLS certificate expiry (HTTPS services) |
waylight_containers |
gauge | state (running, stopped, unhealthy) |
Number of containers |
waylight_container_running |
gauge | container |
1 if the container is running |
waylight_backup_ok |
gauge | name, kind (backup, check) |
1 if ok or running, 0 if overdue or failed (waiting sources are omitted) |
waylight_backup_last_success_timestamp_seconds |
gauge | name, kind |
Last successful run |
waylight_host_cpu_ratio |
gauge | – | CPU usage (0–1) |
waylight_host_load1 |
gauge | – | 1-minute load average |
waylight_host_memory_used_bytes, waylight_host_memory_total_bytes |
gauge | – | Memory |
waylight_host_disk_used_bytes, waylight_host_disk_total_bytes |
gauge | disk |
Per SYSTEM_DISK_PATHS entry |
waylight_host_temperature_celsius |
gauge | – | Highest sensor reading (omitted without sensors) |
waylight_host_uptime_seconds |
gauge | – | Host uptime |
waylight_http_requests_total |
counter | method, code (2xx … 5xx) |
Requests served by Waylight |
waylight_http_request_duration_seconds_sum / _count |
counter | – | Time spent serving requests |
go_goroutines, go_memstats_heap_alloc_bytes, go_memstats_sys_bytes, go_gc_cycles_total |
gauge/counter | – | Go runtime |
process_start_time_seconds |
gauge | – | Start of the process |
Metrics of disabled modules are simply missing.
Design¶
Modules implement the format-neutral kernel.MetricSource port and know
nothing about Prometheus. cmd/server collects the samples of all enabled
modules and internal/platform/metrics renders them by hand (no client
library, see ADR 0010); runtime metrics come from the standard library's
runtime/metrics.