b840047e19
ci: add GitHub Pages deployment workflow for docs site Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
61 lines
2.4 KiB
Go
61 lines
2.4 KiB
Go
// Package observability provides Prometheus metrics and health endpoints for the poller.
|
|
package observability
|
|
|
|
import (
|
|
"github.com/prometheus/client_golang/prometheus"
|
|
"github.com/prometheus/client_golang/prometheus/promauto"
|
|
)
|
|
|
|
// PollDuration tracks the duration of individual device poll cycles.
|
|
var PollDuration = promauto.NewHistogram(prometheus.HistogramOpts{
|
|
Name: "mikrotik_poll_duration_seconds",
|
|
Help: "Duration of a single device poll cycle in seconds.",
|
|
Buckets: []float64{0.5, 1, 2, 5, 10, 30, 60},
|
|
})
|
|
|
|
// PollTotal counts the total number of poll cycles by status.
|
|
// Status labels: "success", "error", "skipped".
|
|
var PollTotal = promauto.NewCounterVec(prometheus.CounterOpts{
|
|
Name: "mikrotik_poll_total",
|
|
Help: "Total number of poll cycles.",
|
|
}, []string{"status"})
|
|
|
|
// DevicesActive tracks the number of devices currently being polled.
|
|
var DevicesActive = promauto.NewGauge(prometheus.GaugeOpts{
|
|
Name: "mikrotik_devices_active",
|
|
Help: "Number of devices currently being polled.",
|
|
})
|
|
|
|
// DeviceConnectionErrors counts total device connection failures.
|
|
var DeviceConnectionErrors = promauto.NewCounter(prometheus.CounterOpts{
|
|
Name: "mikrotik_device_connection_errors_total",
|
|
Help: "Total device connection failures.",
|
|
})
|
|
|
|
// NATSPublishTotal counts NATS publish operations by subject and status.
|
|
// Subject labels: "status", "metrics", "firmware".
|
|
// Status labels: "success", "error".
|
|
var NATSPublishTotal = promauto.NewCounterVec(prometheus.CounterOpts{
|
|
Name: "mikrotik_nats_publish_total",
|
|
Help: "Total NATS publish operations.",
|
|
}, []string{"subject", "status"})
|
|
|
|
// RedisLockTotal counts Redis lock operations by status.
|
|
// Status labels: "obtained", "not_obtained", "error".
|
|
var RedisLockTotal = promauto.NewCounterVec(prometheus.CounterOpts{
|
|
Name: "mikrotik_redis_lock_total",
|
|
Help: "Total Redis lock operations.",
|
|
}, []string{"status"})
|
|
|
|
// CircuitBreakerSkips counts polls skipped due to circuit breaker backoff.
|
|
var CircuitBreakerSkips = promauto.NewCounter(prometheus.CounterOpts{
|
|
Name: "mikrotik_circuit_breaker_skips_total",
|
|
Help: "Total polls skipped because the device is in circuit breaker backoff.",
|
|
})
|
|
|
|
// CircuitBreakerResets counts circuit breaker resets (device recovered after failures).
|
|
var CircuitBreakerResets = promauto.NewCounter(prometheus.CounterOpts{
|
|
Name: "mikrotik_circuit_breaker_resets_total",
|
|
Help: "Total circuit breaker resets when a device recovers.",
|
|
})
|