Struct PrometheusSubscriber 

pub struct PrometheusSubscriber { /* private fields */ }

Prometheus subscriber for supervision-level metrics.

Implements Subscribe and captures metrics from the taskvisor event stream. Must share a Registry with crate::PrometheusMetrics for a unified /metrics endpoint.

Event → metric mapping

TaskStarting        → tasks_in_flight.inc()
                      + task_restarts.inc()  (if attempt > 1)
TaskStopped         → tasks_in_flight.dec()
TaskFailed          → tasks_in_flight.dec()
TimeoutHit          → task_timeouts.inc()
BackoffScheduled    → task_backoff_count{source}.inc()
                      + task_backoff_duration.observe(delay)
ActorExhausted      → task_terminal{reason="exhausted"}.inc()
                      + attempts_to_finalize{outcome="exhausted"}.observe(attempt)
ActorDead           → task_terminal{reason="fatal"}.inc()
                      + attempts_to_finalize{outcome="fatal"}.observe(attempt)
SubscriberOverflow  → subscriber_overflow.inc() + tracing::warn
SubscriberPanicked  → subscriber_panicked.inc() + tracing::warn
ControllerSubmitted → controller_submissions.inc()
ControllerRejected  → controller_rejections{reason}.inc()  (reason classified from Event.reason)

Supervision metrics (solti_sv_*)

Metric	Type	Labels	Description
solti_sv_tasks_in_flight	Gauge	-	Currently executing tasks
solti_sv_task_restarts_total	Counter	-	Restarts (attempt > 1)
solti_sv_task_backoff_count_total	Counter	source	Backoff events
solti_sv_task_backoff_duration_seconds	Histogram	-	Backoff delay duration
solti_sv_task_terminal_total	Counter	reason	Terminal task states
solti_sv_attempts_to_finalize	Histogram	outcome	Attempts when task left loop
solti_sv_task_timeouts_total	Counter	-	Timeout events
solti_sv_subscriber_overflow_total	Counter	-	Queue overflow (lost events)
solti_sv_subscriber_panicked_total	Counter	-	Subscriber panics

Controller metrics (solti_ctrl_*)

Metric	Type	Labels	Description
solti_ctrl_submissions_total	Counter	-	Controller submissions
solti_ctrl_rejections_total	CounterVec	reason	Controller rejections grouped by cause

Labels

Label	Values	Source
source	failure, success	BackoffSource on the event
reason (terminal)	exhausted, fatal	Terminal event kind
outcome (attempts)	exhausted, fatal	Attempts-to-finalize histogram
reason (rejection)	slot_full, slot_busy, add_failed, remove_failed, queue_failed, recovery_failed, bus_lagged, controller_exited, other, unknown	Classified from Event.reason by classify_rejection_reason (private)

Notes

• tasks_in_flight gauge is guarded against going negative: a TaskStopped without a preceding TaskStarting is a no-op on the gauge.
• queue_capacity defaults to DEFAULT_QUEUE_CAPACITY.
• Backoff duration is converted from milliseconds to seconds before observation.

Also

• PrometheusMetrics is a runner-level metrics, complementary to this subscriber.
• Event and EventKind: event structure and classification.

Implementations

impl PrometheusSubscriber

pub fn new(registry: Arc<Registry>) -> Result<Self, Error>

Create a new subscriber with the default event-bus queue capacity (DEFAULT_QUEUE_CAPACITY).

pub fn with_queue_capacity( registry: Arc<Registry>, queue_capacity: usize, ) -> Result<Self, Error>

Create a new subscriber with a specific event-bus queue capacity.

Trait Implementations

impl Debug for PrometheusSubscriber

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Subscribe for PrometheusSubscriber

fn on_event(&self, event: &Event)

Translates a taskvisor event into prometheus metric updates.

fn name(&self) -> &'static str

Returns "prometheus"

fn queue_capacity(&self) -> usize

Returns the per-subscriber queue capacity configured via PrometheusSubscriber::new or PrometheusSubscriber::with_queue_capacity.