Metrics Reference¶

Verified by tests

DispatcherMetricsTests, LifecycleMetricsTests, LifecycleCoordinatorMetricsTests, TransportMetricsTests, PerspectiveMetricsTests, PerspectiveRewindMetricsTests, WorkCoordinatorMetricsTests, InboxMetricsTests, DeadLetterMetricsTests, EventCategoryMetricsTests, TableStatisticsMetricsTests, TypeRegistryMetricsTests, StreamIntegrityMetricsTests, SagaMetricsTests — library CI run #31657041675 (2026-08-13)

Whizbang emits structured OpenTelemetry metrics from every major subsystem. Each subsystem owns a dedicated Meter so you can selectively subscribe to the instruments you care about. All metrics are near-zero cost when no exporter is attached - the .NET System.Diagnostics.Metrics API short-circuits recording when there are no listeners.

The six core meters cover the main request lifecycle:

Meter	Class	Scope
`Whizbang.Dispatcher`	`DispatcherMetrics`	Send, Publish, LocalInvoke, cascade, serialization
`Whizbang.Lifecycle`	`LifecycleMetrics`	Stage execution, receptor invocation, tag hooks
`Whizbang.LifecycleCoordinator`	`LifecycleCoordinatorMetrics`	Coordinator state, WhenAll tracking, stage firing
`Whizbang.Transport`	`TransportMetrics`	Inbox receive, outbox publish, event store, subscriptions, body offloads
`Whizbang.Perspectives`	`PerspectiveMetrics`	Perspective worker batches, checkpoints, event loading, rewinds
`Whizbang.WorkCoordinator`	`WorkCoordinatorMetrics` (+ `InboxMetrics`)	process_work_batch SQL, flush, publisher worker, maintenance, gate holds, inbox dispatch

Additional subsystem meters:

Meter	Class	Scope
`Whizbang.DeadLetters`	`DeadLetterMetrics`	Internal DLQ adds, recoveries, holds, generation replay
`Whizbang.TransportDeadLetterDrain`	`TransportDeadLetterDrainWorker`	Broker-side DLQ drain counts
`Whizbang.EventCategories`	`EventCategoryMetrics`	Category-routed event dispatch and fanout
`Whizbang.Workers.PinnedPool`	`PinnedPoolMetrics`	Pinned connection pool borrows, timeouts, recycles
`Whizbang.TableStatistics`	`TableStatisticsMetrics`	Estimated queue depth and table size gauges
`Whizbang.TypeRegistry`	`TypeRegistryMetrics`	Message type registry renames and drift
`Whizbang.StreamIntegrity`	`StreamIntegrityMetrics`	Continuity checkpoints, gap/divergence detection, repair and manifest flows
`Whizbang.Core.Routing.MessageDiscard`	`MessageDiscardPolicy`	Unsubscribed-message discards at the receive boundary
`Whizbang.Postgres.Notifications`	`NotifyMetrics`	LISTEN/NOTIFY signal delivery and connection state
`Whizbang.Sagas`	`SagaMetrics`	Saga initiation, completion, item and hook outcomes

Configuration¶

All metrics classes are automatically registered as singletons by AddWhizbang(). The shared WhizbangMetrics class holds the IMeterFactory reference that each subsystem uses to create its meter.

Metrics Registration

// Metrics are registered automatically - no opt-in required
services.AddWhizbang(options => {
  // ... your configuration
});

// To export metrics, configure OpenTelemetry SDK
builder.Services.AddOpenTelemetry()
    .WithMetrics(metrics => {
      // Subscribe to all Whizbang meters
      metrics.AddMeter("Whizbang.Dispatcher");
      metrics.AddMeter("Whizbang.Lifecycle");
      metrics.AddMeter("Whizbang.LifecycleCoordinator");
      metrics.AddMeter("Whizbang.Transport");
      metrics.AddMeter("Whizbang.Perspectives");
      metrics.AddMeter("Whizbang.WorkCoordinator");

      // Optional subsystem meters (subscribe to the ones you use)
      metrics.AddMeter("Whizbang.DeadLetters");
      metrics.AddMeter("Whizbang.TransportDeadLetterDrain");
      metrics.AddMeter("Whizbang.EventCategories");
      metrics.AddMeter("Whizbang.Workers.PinnedPool");
      metrics.AddMeter("Whizbang.TableStatistics");
      metrics.AddMeter("Whizbang.TypeRegistry");
      metrics.AddMeter("Whizbang.StreamIntegrity");
      metrics.AddMeter("Whizbang.Core.Routing.MessageDiscard");
      metrics.AddMeter("Whizbang.Postgres.Notifications");
      metrics.AddMeter("Whizbang.Sagas");

      // Export to Prometheus, OTLP, or Aspire
      metrics.AddPrometheusExporter();
      // or: metrics.AddOtlpExporter();
    });

WhizbangMetrics¶

The WhizbangMetrics class is the shared parent that owns the IMeterFactory reference. All subsystem metrics classes inject it to create their meters. When IMeterFactory is available (e.g., via OpenTelemetry SDK registration), meters are created through the factory. Otherwise, a standalone Meter is created as a fallback.

WhizbangMetrics

// Registered automatically as singleton
public sealed class WhizbangMetrics(IMeterFactory? meterFactory = null) {
  public IMeterFactory? MeterFactory { get; } = meterFactory;
}

Whizbang.Dispatcher¶

Meter name: Whizbang.Dispatcher

Instruments covering the full dispatch path - from SendAsync entry through receptor invocation, cascade extraction, serialization, and perspective synchronization.

Timing Histograms¶

Metric Name	Unit	Description
`whizbang.dispatcher.send.duration`	ms	SendAsync: envelope creation, receptor, cascade, lifecycle
`whizbang.dispatcher.publish.duration`	ms	PublishAsync: local handlers, outbox queue, flush
`whizbang.dispatcher.local_invoke.duration`	ms	LocalInvokeAsync: receptor invocation only
`whizbang.dispatcher.local_invoke_and_sync.duration`	ms	LocalInvokeAndSyncAsync: invoke + perspective wait
`whizbang.dispatcher.cascade.duration`	ms	CascadeMessageAsync: extraction, routing, outbox/local
`whizbang.dispatcher.send_many.duration`	ms	SendManyAsync: batch dispatch total

Sub-Operation Histograms¶

Metric Name	Unit	Description
`whizbang.dispatcher.receptor.duration`	ms	Time in receptor delegate invocation
`whizbang.dispatcher.cascade_extraction.duration`	ms	MessageExtractor.ExtractMessagesWithRouting
`whizbang.dispatcher.perspective_sync.duration`	ms	_awaitPerspectiveSyncIfNeededAsync wait time
`whizbang.dispatcher.perspective_wait.duration`	ms	_waitForPerspectivesIfNeededAsync (post-dispatch)
`whizbang.dispatcher.serialization.duration`	ms	_serializeToNewOutboxMessage JSON serialization
`whizbang.dispatcher.tag_processing.duration`	ms	_processTagsIfEnabledAsync execution

Throughput Counters¶

Metric Name	Type	Description
`whizbang.dispatcher.messages_dispatched`	Counter\<long>	Total messages dispatched
`whizbang.dispatcher.events_cascaded`	Counter\<long>	Events cascaded from receptor results
`whizbang.dispatcher.messages_serialized`	Counter\<long>	Messages serialized for outbox
`whizbang.dispatcher.duplicates_detected`	Counter\<long>	Inbox dedup rejections
`whizbang.dispatcher.perspective_sync_timeouts`	Counter\<long>	Perspective sync wait timeouts
`whizbang.dispatcher.errors`	Counter\<long>	Dispatch-level errors
`whizbang.dispatcher.publish_once.claims_won`	Counter\<long>	PublishOnceAsync calls that won the claim and emitted the event
`whizbang.dispatcher.publish_once.claims_lost`	Counter\<long>	PublishOnceAsync calls that lost the claim and intentionally no-opped

Batch Histograms¶

Metric Name	Type	Description
`whizbang.dispatcher.cascade.event_count`	Histogram\<int>	Events extracted per cascade
`whizbang.dispatcher.send_many.batch_size`	Histogram\<int>	Messages per SendMany call

Whizbang.Lifecycle¶

Meter name: Whizbang.Lifecycle

Instruments for all 20 lifecycle stages, individual receptor invocations, and the tag hook pipeline.

Stage Timing¶

Metric Name	Unit	Description
`whizbang.lifecycle.stage.duration`	ms	Time executing all receptors for a stage
`whizbang.lifecycle.receptor.duration`	ms	Individual receptor invocation time

Stage Counters¶

Metric Name	Type	Description
`whizbang.lifecycle.stage.invocations`	Counter\<long>	Total invocations per lifecycle stage
`whizbang.lifecycle.receptor.invocations`	Counter\<long>	Individual receptor invocations
`whizbang.lifecycle.receptor.errors`	Counter\<long>	Receptor failures per stage

Tag Hook Timing¶

Metric Name	Unit	Description
`whizbang.lifecycle.tag_hook.duration`	ms	Per-hook execution time
`whizbang.lifecycle.tag_processing.duration`	ms	Total tag processing time (all hooks)

Tag Hook Counters¶

Metric Name	Type	Description
`whizbang.lifecycle.tag_hook.invocations`	Counter\<long>	Hook invocations
`whizbang.lifecycle.tag_hook.errors`	Counter\<long>	Hook failures

Whizbang.LifecycleCoordinator¶

Meter name: Whizbang.LifecycleCoordinator

Instruments for the coordinator that tracks active events through the lifecycle, manages perspective WhenAll gates, and fires PostAllPerspectives/PostLifecycle stages.

Active Tracking Gauges¶

Metric Name	Type	Description
`whizbang.lifecycle_coordinator.active_tracked_events`	UpDownCounter\<int>	Events currently in lifecycle tracking
`whizbang.lifecycle_coordinator.pending_perspective_states`	UpDownCounter\<int>	Events awaiting perspective WhenAll completion
`whizbang.lifecycle_coordinator.pending_when_all_states`	UpDownCounter\<int>	Events awaiting segment WhenAll completion

Completion Counters¶

Metric Name	Type	Description
`whizbang.lifecycle_coordinator.perspective_completions_signaled`	Counter\<long>	Individual perspective complete signals received
`whizbang.lifecycle_coordinator.all_perspectives_completed`	Counter\<long>	Events where all perspectives finished
`whizbang.lifecycle_coordinator.expectations_not_registered`	Counter\<long>	Events with no perspective expectations (key mismatch detector)

Stage Firing Counters¶

Metric Name	Type	Description
`whizbang.lifecycle_coordinator.post_all_perspectives_fired`	Counter\<long>	PostAllPerspectives stage executions
`whizbang.lifecycle_coordinator.post_lifecycle_fired`	Counter\<long>	PostLifecycle stage executions
`whizbang.lifecycle_coordinator.post_lifecycle_errors`	Counter\<long>	PostLifecycle stage failures
`whizbang.lifecycle_coordinator.stage_transitions`	Counter\<long>	Stage transitions (tag: stage)

Cleanup Counters¶

Metric Name	Type	Description
`whizbang.lifecycle_coordinator.stale_tracking_cleaned`	Counter\<long>	Stale tracking entries cleaned by inactivity threshold
`whizbang.lifecycle_coordinator.stale_tracking_preserved_partial_perspectives`	Counter\<long>	Stale entries preserved because some perspectives had already completed

Whizbang.Transport¶

Meter name: Whizbang.Transport

Instruments covering the inbox receive pipeline, outbox publish pipeline, transport subscriptions, and event store operations.

Inbox Timing¶

Metric Name	Unit	Description
`whizbang.transport.inbox.receive.duration`	ms	Full _handleMessageAsync: receive, process, complete
`whizbang.transport.inbox.dedup.duration`	ms	First FlushAsync (INSERT ... ON CONFLICT)
`whizbang.transport.inbox.processing.duration`	ms	OrderedStreamProcessor.ProcessInboxWorkAsync
`whizbang.transport.inbox.completion.duration`	ms	Second FlushAsync (report completions)
`whizbang.transport.inbox.security_context.duration`	ms	SecurityContextHelper.EstablishFullContextAsync
`whizbang.transport.inbox.concurrency_wait.duration`	ms	Time waiting for a concurrency semaphore slot
`whizbang.transport.inbox.batch.wait.duration`	ms	Time the first message in a batch waited before flush

Inbox Counters¶

Metric Name	Type	Description
`whizbang.transport.inbox.messages_received`	Counter\<long>	Messages received from transport
`whizbang.transport.inbox.messages_processed`	Counter\<long>	Successfully processed
`whizbang.transport.inbox.messages_deduplicated`	Counter\<long>	Rejected as duplicates
`whizbang.transport.inbox.messages_failed`	Counter\<long>	Processing failures
`whizbang.transport.inbox.subscription_retries`	Counter\<long>	Transport subscription retry attempts
`whizbang.transport.inbox.batch.flushes`	Counter\<long>	Total inbox batch flushes
`whizbang.transport.inbox.batch.size`	Histogram\<double>	Messages per inbox batch flush
`whizbang.transport.inbox.concurrent_messages`	UpDownCounter\<int>	Current concurrent message handlers

Outbox Timing¶

Metric Name	Unit	Description
`whizbang.transport.outbox.publish.duration`	ms	_publishStrategy.PublishAsync to transport
`whizbang.transport.outbox.readiness_wait.duration`	ms	Time waiting for transport readiness

Outbox Counters¶

Metric Name	Type	Description
`whizbang.transport.outbox.messages_published`	Counter\<long>	Messages published to transport
`whizbang.transport.outbox.messages_failed`	Counter\<long>	Publish failures by reason
`whizbang.transport.outbox.publish_retries`	Counter\<long>	Retry attempts
`whizbang.transport.outbox.publish_throttled`	Counter\<long>	Broker-side throttle events observed during publish; tagged by `transport`

Subscription Gauges¶

Metric Name	Type	Description
`whizbang.transport.active_subscriptions`	UpDownCounter\<int>	Currently active transport subscriptions

Event Store¶

Metric Name	Unit / Type	Description
`whizbang.event_store.append.duration`	ms	AppendAsync latency
`whizbang.event_store.query.duration`	ms	GetEventsBetweenPolymorphicAsync latency
`whizbang.event_store.events_stored`	Counter\<long>	Events appended
`whizbang.event_store.events_queried`	Counter\<long>	Event queries executed

Body Offload (Claim-Check)¶

Metric Name	Unit / Type	Description
`whizbang.transport.body_offload.count`	Counter\<long>	Messages whose body was offloaded (claim-check tripped); tagged by `message.type` + `message.namespace`
`whizbang.transport.body_offload.bytes`	Histogram\<long> (By)	Original serialized size that triggered an offload
`whizbang.transport.body_claim.rehydrated.count`	Counter\<long>	Claim envelopes rehydrated on receive; tagged by `message.type` + `message.namespace`
`whizbang.transport.body_claim.rehydrated.bytes`	Histogram\<long> (By)	Rehydrated body size downloaded from the body store

Whizbang.Perspectives¶

Meter name: Whizbang.Perspectives

Instruments for the perspective worker pipeline - batch processing, claim, event loading, runner execution, and checkpointing.

Timing Histograms¶

Metric Name	Unit	Description
`whizbang.perspective.batch.duration`	ms	Full _processWorkBatchAsync cycle
`whizbang.perspective.claim.duration`	ms	ProcessWorkBatchAsync to claim perspective work
`whizbang.perspective.event_load.duration`	ms	GetEventsBetweenPolymorphicAsync query
`whizbang.perspective.runner.duration`	ms	IPerspectiveRunner execution per stream
`whizbang.perspective.checkpoint.duration`	ms	GetPerspectiveCursorAsync

Throughput Counters¶

Metric Name	Type	Description
`whizbang.perspective.events_processed`	Counter\<long>	Events applied to perspectives
`whizbang.perspective.batches_processed`	Counter\<long>	Batches completed
`whizbang.perspective.streams_updated`	Counter\<long>	Unique streams updated
`whizbang.perspective.errors`	Counter\<long>	Processing errors
`whizbang.perspective.empty_batches`	Counter\<long>	Polling cycles with no work

Backlog & Rewind¶

Metric Name	Unit / Type	Description
`whizbang.perspective.pending_events`	ObservableGauge	Pending perspective events awaiting processing
`whizbang.perspective.rewinds`	Counter\<long>	Rewind operations triggered
`whizbang.perspective.rewind.duration`	ms	Rewind replay duration
`whizbang.perspective.rewind.events_replayed`	Histogram\<int>	Events replayed per rewind
`whizbang.perspective.rewind.events_behind`	Histogram\<int>	Events behind cursor when rewind triggered

Batch Composition Histograms¶

Metric Name	Type	Description
`whizbang.perspective.batch.work_items`	Histogram\<int>	Work items claimed per batch
`whizbang.perspective.batch.event_count`	Histogram\<int>	Events loaded per batch
`whizbang.perspective.batch.stream_groups`	Histogram\<int>	Distinct streams per batch

Whizbang.WorkCoordinator¶

Meter name: Whizbang.WorkCoordinator

Instruments for the core work coordination pipeline - the process_work_batch SQL call, flush orchestration, publisher worker, and maintenance tasks.

Timing Histograms¶

Metric Name	Unit	Description
`whizbang.work_coordinator.process_batch.duration`	ms	Time executing process_work_batch SQL
`whizbang.work_coordinator.flush.duration`	ms	Total FlushAsync time including lifecycle

Batch Composition (Input)¶

Metric Name	Type	Description
`whizbang.work_coordinator.batch.outbox_messages`	Histogram\<int>	Outbox messages sent to process_work_batch
`whizbang.work_coordinator.batch.inbox_messages`	Histogram\<int>	Inbox messages sent
`whizbang.work_coordinator.batch.completions`	Histogram\<int>	Completions sent
`whizbang.work_coordinator.batch.failures`	Histogram\<int>	Failures sent

Work Returned (Output)¶

Metric Name	Type	Description
`whizbang.work_coordinator.returned.outbox_work`	Histogram\<int>	Outbox work items returned
`whizbang.work_coordinator.returned.inbox_work`	Histogram\<int>	Inbox work items returned
`whizbang.work_coordinator.returned.perspective_work`	Histogram\<int>	Perspective work items returned

Counters¶

Metric Name	Type	Description
`whizbang.work_coordinator.process_batch.calls`	Counter\<long>	Total process_work_batch calls
`whizbang.work_coordinator.process_batch.errors`	Counter\<long>	SQL errors
`whizbang.work_coordinator.flush.calls`	Counter\<long>	Total FlushAsync calls
`whizbang.work_coordinator.flush.empty_calls`	Counter\<long>	Flushes with no queued work

Publisher Worker¶

Metric Name	Type	Description
`whizbang.publisher.lease_renewals`	Counter\<long>	Lease renewals due to transport not ready
`whizbang.publisher.buffered_messages`	Counter\<long>	Total messages buffered for publish

Maintenance¶

Metric Name	Unit / Type	Description
`whizbang.maintenance.task.duration`	ms	Duration per maintenance task
`whizbang.maintenance.task.rows_affected`	Histogram\<long>	Rows cleaned per task

Gate & Inbox Dispatch¶

Both instruments live on the Whizbang.WorkCoordinator meter (InboxMetrics deliberately reuses it):

Metric Name	Unit / Type	Description
`whizbang.gate.hold_duration_ms`	ms	WorkCoordinatorGate slot-held duration; tagged with caller
`whizbang.inbox.dispatch.duration_ms`	ms	Per-message inbox dispatch wall time, tagged with short message type

Whizbang.DeadLetters¶

Meter name: Whizbang.DeadLetters (DeadLetterMetrics)

Metric Name	Type	Description
`whizbang.dead_letters.added`	Counter\<long>	Rows moved into `wh_dead_letters`; tagged by `source_table` + `reason`
`whizbang.dead_letters.recovered`	Counter\<long>	Successful recovery re-emits; tagged by `source_table`
`whizbang.dead_letters.held`	Counter\<long>	Transitions to HoldForReview; tagged by `policy_name` + `reason`
`whizbang.dead_letters.permanently_failed`	Counter\<long>	Transitions to PermanentlyFailed; tagged by `policy_name` + `reason`
`whizbang.dead_letters.recovery_attempts`	Counter\<long>	Recovery attempts dispatched (any outcome); tagged by `reason`
`whizbang.dead_letters.generation_replay_scheduled`	Counter\<long>	Rows scheduled by the generation-replay sweep; tagged by `generation`

Whizbang.TransportDeadLetterDrain¶

Meter name: Whizbang.TransportDeadLetterDrain (created by TransportDeadLetterDrainWorker)

Metric Name	Type	Description
`whizbang.transport_dlq.drained`	Counter\<long>	Messages re-submitted from a broker's dead-letter queue; tagged by `transport`

Whizbang.EventCategories¶

Meter name: Whizbang.EventCategories (EventCategoryMetrics)

Metric Name	Unit / Type	Description
`whizbang.event_category.dispatch.duration`	ms	Dispatch / expansion duration at the category seam (collective dispatcher or composite expander)
`whizbang.event_category.fanout`	Histogram\<int>	Fan-out factor — affected rows for collective, inner-event count for composite
`whizbang.event_category.dispatched`	Counter\<long>	Dispatches per category
`whizbang.event_category.errors`	Counter\<long>	Dispatch failures per category

Whizbang.Workers.PinnedPool¶

Meter name: Whizbang.Workers.PinnedPool (PinnedPoolMetrics)

Metric Name	Unit / Type	Description
`whizbang.workers.pinned_pool.borrow.duration`	ms	Time from borrow request to connection handed back by the pinned pool
`whizbang.workers.pinned_pool.borrow.timeouts`	Counter\<long>	Borrow attempts that timed out waiting for an available connection
`whizbang.workers.pinned_pool.connection_recycles`	Counter\<long>	Pinned-connection recycles (Npgsql ConnectionLifetime)

Whizbang.TableStatistics¶

Meter name: Whizbang.TableStatistics (TableStatisticsMetrics)

Metric Name	Type	Description
`whizbang.queue.estimated_depth`	ObservableGauge	Unprocessed message count for inbox/outbox queues
`whizbang.table.estimated_bytes`	ObservableGauge	Estimated disk size per table from the database catalog

Whizbang.TypeRegistry¶

Meter name: Whizbang.TypeRegistry (TypeRegistryMetrics)

Metric Name	Type	Description
`whizbang.type_registry.renamed`	Counter\<long>	Registry rows reconciled old→new for an acknowledged rename; tagged by `service`
`whizbang.type_registry.drift_detected`	Counter\<long>	Un-acknowledged registry drift left untouched; tagged by `service`

Whizbang.StreamIntegrity¶

Meter name: Whizbang.StreamIntegrity (StreamIntegrityMetrics)

Metric Name	Type	Description
`whizbang.stream_integrity.checkpoints_published`	Counter\<long>	Continuity checkpoints published (empty windows included - the liveness beat)
`whizbang.stream_integrity.checkpoints_received`	Counter\<long>	Continuity checkpoints received; tagged by `origin`
`whizbang.stream_integrity.gaps_detected`	Counter\<long>	Confirmed continuity gaps; tagged by `origin` + `event_type`
`whizbang.stream_integrity.divergences_detected`	Counter\<long>	Confirmed audit divergences; tagged by `origin` + `event_type`
`whizbang.stream_integrity.coverage_gaps_detected`	Counter\<long>	Local perspective coverage gaps; tagged by `perspective`
`whizbang.stream_integrity.repairs_requested`	Counter\<long>	Scoped re-delivery repair requests sent; tagged by `source` (checkpoint \| audit) + `origin`
`whizbang.stream_integrity.rebuilds_requested`	Counter\<long>	Local rebuilds dispatched for coverage gaps; tagged by `perspective`
`whizbang.stream_integrity.manifests_requested`	Counter\<long>	Manifest requests sent to origins; tagged by `origin` + `sweep`
`whizbang.stream_integrity.manifest_chunks_sent`	Counter\<long>	Manifest chunks answered as an origin; tagged by `level`
`whizbang.stream_integrity.drill_downs_requested`	Counter\<long>	Type-level mismatches escalated to stream-level requests; tagged by `origin`
`whizbang.stream_integrity.backfills_requested`	Counter\<long>	State-only backfill requests broadcast for consumed-set growth (value = new type count)
`whizbang.stream_integrity.digest_buckets_verified`	Counter\<long>	Settled digest buckets checked by the trust-but-verify sweep
`whizbang.stream_integrity.digest_drift_healed`	Counter\<long>	Digest buckets the sweep healed; tagged by `kind` (updated \| removed \| added)
`whizbang.stream_integrity.redelivery_requests_received`	Counter\<long>	Re-delivery requests served as an origin (repair + backfill flows)

Other Meters¶

Meter	Metric Name	Type	Description
`Whizbang.Core.Routing.MessageDiscard`	`whizbang.message.skipped`	Counter\<long>	Messages intentionally skipped at a discard gate (receive \| inbox \| outbox)
`Whizbang.Postgres.Notifications`	`whizbang.postgres.notifications.signals_received`	Counter\<long>	NOTIFY signals delivered to subscribers; tagged by category (outbox/inbox/perspective/unknown)
`Whizbang.Postgres.Notifications`	`whizbang.postgres.notifications.connection_state`	UpDownCounter\<int>	+1 when LISTEN/NOTIFY is healthy, -1 on disconnect; sum = number of connected pods
`Whizbang.Postgres.Notifications`	`whizbang.postgres.notifications.signaling_mode`	Counter\<long>	Mode-decision events (startup + runtime transitions); tagged by mode and reason
`Whizbang.Sagas`	`whizbang.sagas.initiated`	Counter\<long>	Sagas initiated
`Whizbang.Sagas`	`whizbang.sagas.completed`	Counter\<long>	Sagas reaching a terminal completed state
`Whizbang.Sagas`	`whizbang.sagas.failed`	Counter\<long>	Sagas that fail-fasted
`Whizbang.Sagas`	`whizbang.sagas.duration`	Histogram\<double> (s)	End-to-end saga duration
`Whizbang.Sagas`	`whizbang.sagas.items_completed`	Histogram\<int>	Items completed per saga
`Whizbang.Sagas`	`whizbang.sagas.items_failed`	Histogram\<int>	Items failed per saga
`Whizbang.Sagas`	`whizbang.sagas.items_reset`	Counter\<long>	Saga items reset via SagaResetEvent
`Whizbang.Sagas`	`whizbang.sagas.hooks_completed`	Counter\<long>	Saga hooks that succeeded
`Whizbang.Sagas`	`whizbang.sagas.hooks_failed`	Counter\<long>	Saga hooks that failed

Grafana and Prometheus Integration¶

Whizbang metrics are standard OpenTelemetry instruments, making them compatible with any OTel-aware backend. The most common production setup is Prometheus scraping with Grafana dashboards.

Prometheus Exporter¶

Prometheus Setup

builder.Services.AddOpenTelemetry()
    .WithMetrics(metrics => {
      metrics.AddMeter("Whizbang.Dispatcher");
      metrics.AddMeter("Whizbang.Lifecycle");
      metrics.AddMeter("Whizbang.LifecycleCoordinator");
      metrics.AddMeter("Whizbang.Transport");
      metrics.AddMeter("Whizbang.Perspectives");
      metrics.AddMeter("Whizbang.WorkCoordinator");
      metrics.AddPrometheusExporter();
    });

// Expose /metrics endpoint
app.MapPrometheusScrapingEndpoint();

Useful Grafana Queries¶

Grafana PromQL Examples

# Dispatch throughput (messages/sec)
rate(whizbang_dispatcher_messages_dispatched_total[5m])

# p99 send latency
histogram_quantile(0.99, rate(whizbang_dispatcher_send_duration_bucket[5m]))

# Inbox processing error rate
rate(whizbang_transport_inbox_messages_failed_total[5m])
  / rate(whizbang_transport_inbox_messages_received_total[5m])

# Active events in lifecycle coordinator (gauge)
whizbang_lifecycle_coordinator_active_tracked_events

# Perspective lag (empty batches indicate worker is caught up)
rate(whizbang_perspective_empty_batches_total[5m])

# process_work_batch SQL p95 latency
histogram_quantile(0.95, rate(whizbang_work_coordinator_process_batch_duration_bucket[5m]))

Aspire Dashboard¶

For .NET Aspire projects, metrics appear automatically in the Aspire dashboard:

Aspire Metrics Integration

// In ServiceDefaults project
builder.Services.AddOpenTelemetry()
    .WithMetrics(metrics => {
      metrics.AddMeter("Whizbang.Dispatcher");
      metrics.AddMeter("Whizbang.Lifecycle");
      metrics.AddMeter("Whizbang.LifecycleCoordinator");
      metrics.AddMeter("Whizbang.Transport");
      metrics.AddMeter("Whizbang.Perspectives");
      metrics.AddMeter("Whizbang.WorkCoordinator");
    });

Instrument Types¶

Whizbang uses four OpenTelemetry instrument types:

Instrument	Purpose	Example
`Counter<T>`	Monotonically increasing totals	`messages_dispatched`, `errors`
`Histogram<T>`	Value distributions (latency, batch sizes)	`send.duration`, `batch.event_count`
`UpDownCounter<T>`	Gauges that can increase or decrease	`active_tracked_events`, `active_subscriptions`
`ObservableGauge`	Pull-based point-in-time values	`pending_events`, `queue.estimated_depth`

Histograms are ideal for latency percentiles (p50, p95, p99) and batch size distributions. Counters track cumulative totals - use rate() in Prometheus to derive per-second throughput. UpDownCounters represent current state and are useful for alerting on resource saturation.

Metrics Reference¶

Configuration¶

WhizbangMetrics¶

Whizbang.Dispatcher¶

Timing Histograms¶

Sub-Operation Histograms¶

Throughput Counters¶

Batch Histograms¶

Whizbang.Lifecycle¶

Stage Timing¶

Stage Counters¶

Tag Hook Timing¶

Tag Hook Counters¶

Whizbang.LifecycleCoordinator¶

Active Tracking Gauges¶

Completion Counters¶

Stage Firing Counters¶

Cleanup Counters¶

Whizbang.Transport¶

Inbox Timing¶

Inbox Counters¶

Outbox Timing¶

Outbox Counters¶

Subscription Gauges¶

Event Store¶

Body Offload (Claim-Check)¶

Whizbang.Perspectives¶

Timing Histograms¶

Throughput Counters¶

Backlog & Rewind¶

Batch Composition Histograms¶

Whizbang.WorkCoordinator¶

Timing Histograms¶

Batch Composition (Input)¶

Work Returned (Output)¶

Counters¶

Publisher Worker¶

Maintenance¶

Gate & Inbox Dispatch¶

Whizbang.DeadLetters¶

Whizbang.TransportDeadLetterDrain¶

Whizbang.EventCategories¶

Whizbang.Workers.PinnedPool¶

Whizbang.TableStatistics¶

Whizbang.TypeRegistry¶

Whizbang.StreamIntegrity¶

Other Meters¶

Grafana and Prometheus Integration¶

Prometheus Exporter¶

Useful Grafana Queries¶

Aspire Dashboard¶

Instrument Types¶

See Also¶