metrics

var (
	// AgentErrorsGauge tracks the number of agents in error status per namespace/project
	AgentErrorsGauge = prometheus.NewGaugeVec(
		prometheus.GaugeOpts{
			Name: "flecto_agent_errors_total",
			Help: "Number of agents in error status (excluding offline agents)",
		},
		[]string{"namespace", "project"},
	)

	// AgentOnlineGauge tracks the number of online agents per namespace/project
	AgentOnlineGauge = prometheus.NewGaugeVec(
		prometheus.GaugeOpts{
			Name: "flecto_agent_online_total",
			Help: "Number of online agents",
		},
		[]string{"namespace", "project"},
	)

	// HTTPRequestsTotal counts HTTP requests
	HTTPRequestsTotal = prometheus.NewCounterVec(
		prometheus.CounterOpts{
			Name: "flecto_http_requests_total",
			Help: "Total number of HTTP requests",
		},
		[]string{"method", "path", "status"},
	)

	// HTTPRequestDuration tracks HTTP request duration
	HTTPRequestDuration = prometheus.NewHistogramVec(
		prometheus.HistogramOpts{
			Name:    "flecto_http_request_duration_seconds",
			Help:    "HTTP request duration in seconds",
			Buckets: prometheus.DefBuckets,
		},
		[]string{"method", "path"},
	)

	// SchedulerTaskRunsTotal counts background task runs by outcome. A panic counts
	// as an error, so a task crashing every run never reads as healthy.
	SchedulerTaskRunsTotal = prometheus.NewCounterVec(
		prometheus.CounterOpts{
			Name: "flecto_scheduler_task_runs_total",
			Help: "Total number of background task runs, by outcome",
		},
		[]string{"task", "status"},
	)

	// SchedulerTaskDuration tracks how long a background task run takes, which is
	// how a task drifting towards its timeout becomes visible before it hits it.
	// Buckets span sub-second to a quarter of an hour: these are batch jobs, not
	// requests.
	SchedulerTaskDuration = prometheus.NewHistogramVec(
		prometheus.HistogramOpts{
			Name:    "flecto_scheduler_task_duration_seconds",
			Help:    "Background task run duration in seconds",
			Buckets: []float64{0.1, 0.5, 1, 5, 15, 60, 300, 900},
		},
		[]string{"task"},
	)

	// SchedulerTaskLastSuccessTimestamp is when a background task last succeeded.
	// It catches what a run counter cannot: a task that stopped running altogether
	// emits no error, it simply stops emitting - and its last success ages.
	SchedulerTaskLastSuccessTimestamp = prometheus.NewGaugeVec(
		prometheus.GaugeOpts{
			Name: "flecto_scheduler_task_last_success_timestamp_seconds",
			Help: "Unix timestamp of the last successful run of a background task",
		},
		[]string{"task"},
	)
)