Application Metrics Reference

Every Granit module exposes business-level metrics via the .NET System.Diagnostics.Metrics API. These metrics are automatically exported to Mimir (or any OpenTelemetry-compatible backend) by AddGranitObservability() — no manual wiring needed.

Conventions

All Granit metrics follow strict naming and tagging conventions:

Rule	Convention
File location	src/Granit.{Module}/Diagnostics/{Module}Metrics.cs
Class	sealed class {Module}Metrics with IMeterFactory constructor injection
Meter name	Granit.{Module} (PascalCase, one per module)
Metric name	granit.{module}.{entity}.{action} (all lowercase, dot-separated)
Tags	snake_case, always include tenant_id (coalesced to "global" when null)
Tag passing	TagList (zero-allocation struct) — never KeyValuePair[]
DI registration	services.TryAddSingleton<{Module}Metrics>() in AddGranit*()
Duration unit	seconds ("s"), measured via Stopwatch.GetElapsedTime()

Instrument types

.NET type	OpenTelemetry kind	Use case
Counter<long>	Monotonic sum	Events counted (requests, deliveries, errors)
Histogram<double>	Distribution	Latency, duration
UpDownCounter<long>	Non-monotonic sum	In-flight operations (active leases)

Writing a metrics class

Metrics class

namespace Granit.MyModule.Diagnostics;

public sealed class MyModuleMetrics(IMeterFactory meterFactory)
{
    private readonly Meter _meter = meterFactory.Create("Granit.MyModule");

    private readonly Counter<long> _operationsCompleted;
    private readonly Histogram<double> _operationDuration;

    // Initialize instruments in a method or constructor body
    // ...

    public void RecordOperationCompleted(Guid? tenantId, string status)
    {
        var tags = new TagList
        {
            { "tenant_id", tenantId?.ToString() ?? "global" },
            { "status", status },
        };
        _operationsCompleted.Add(1, tags);
    }

    public void RecordOperationDuration(Guid? tenantId, double seconds)
    {
        var tags = new TagList
        {
            { "tenant_id", tenantId?.ToString() ?? "global" },
        };
        _operationDuration.Record(seconds, tags);
    }
}

DI registration

public static IServiceCollection AddGranitMyModule(this IServiceCollection services)
{
    services.TryAddSingleton<MyModuleMetrics>();
    return services;
}

Unit test

public sealed class MyModuleMetricsTests : IDisposable
{
    private readonly ServiceProvider _sp;
    private readonly MyModuleMetrics _metrics;
    private readonly IMeterFactory _meterFactory;

    public MyModuleMetricsTests()
    {
        _sp = new ServiceCollection().AddMetrics().BuildServiceProvider();
        _meterFactory = _sp.GetRequiredService<IMeterFactory>();
        _metrics = new MyModuleMetrics(_meterFactory);
    }

    [Fact]
    public void RecordOperationCompleted_IncrementsCounter()
    {
        var collector = new MetricCollector<long>(
            _meterFactory, "Granit.MyModule", "granit.mymodule.operations.completed");

        _metrics.RecordOperationCompleted(Guid.NewGuid(), "success");

        var measurement = collector.GetMeasurementSnapshot().Single();
        measurement.Value.ShouldBe(1);
        measurement.Tags["status"].ShouldBe("success");
    }

    public void Dispose() => _sp.Dispose();
}

Metrics by module

AI

Meter: Granit.AI

Metric	Type	Tags	Description
granit.ai.requests.completed	Counter	tenant_id, model, provider, status	LLM requests completed
granit.ai.tokens.input	Counter	tenant_id, model, provider	Input tokens consumed
granit.ai.tokens.output	Counter	tenant_id, model, provider	Output tokens generated
granit.ai.request.duration	Histogram (s)	tenant_id, model, provider	LLM request duration

Auditing

Meter: Granit.Auditing

Metric	Type	Tags	Description
granit.auditing.entries.persisted	Counter	tenant_id, entity_type	Audit entries written
granit.auditing.entries.purged	Counter	tenant_id	Audit entries purged (retention)
granit.auditing.capture.errors	Counter	tenant_id, entity_type	Capture failures

BackgroundJobs

Meter: Granit.BackgroundJobs

Metric	Type	Tags	Description
granit.background_jobs.execution.completed	Counter	tenant_id, job_name, status	Job executions completed
granit.background_jobs.execution.duration	Histogram (s)	tenant_id, job_name, status	Job execution duration

BlobStorage

Meter: Granit.BlobStorage

Metric	Type	Tags	Description
granit.blob_storage.uploads.initiated	Counter	tenant_id, provider	Upload operations started
granit.blob_storage.validations.completed	Counter	tenant_id, status	Blob validations completed
granit.blob_storage.validations.failed	Counter	tenant_id, reason	Blob validations failed
granit.blob_storage.blobs.deleted	Counter	tenant_id	Blobs deleted
granit.blob_storage.orphans.cleaned	Counter	tenant_id	Orphan blobs cleaned up
granit.blob_storage.confirm.duration	Histogram (s)	tenant_id, provider	Blob confirmation duration

DataExchange

Meter: Granit.DataExchange

Metric	Type	Tags	Description
granit.data_exchange.import.jobs.completed	Counter	tenant_id, status	Import jobs completed
granit.data_exchange.import.rows.processed	Counter	tenant_id	Rows processed during import
granit.data_exchange.import.rows.errors	Counter	tenant_id	Rows that failed during import
granit.data_exchange.import.duration	Histogram (s)	tenant_id	Import job duration
granit.data_exchange.export.jobs.completed	Counter	tenant_id, status	Export jobs completed
granit.data_exchange.export.rows.exported	Counter	tenant_id	Rows exported
granit.data_exchange.export.duration	Histogram (s)	tenant_id	Export job duration

Events

Meter: Granit.Events

Metric	Type	Tags	Description
granit.events.event.published	Counter	tenant_id, bus_type, event_type	Events published (local or distributed)
granit.events.handler.executed	Counter	tenant_id, event_type, status	Event handler executions

Identity

Meter: Granit.Identity

Metric	Type	Tags	Description
granit.identity.operations.completed	Counter	tenant_id, operation, provider, status	Identity operations completed
granit.identity.operations.errors	Counter	tenant_id, operation, provider	Identity operation errors
granit.identity.operation.duration	Histogram (s)	tenant_id, operation, provider	Identity operation duration

Notifications

Meter: Granit.Notifications

Metric	Type	Tags	Description
granit.notifications.fanout.triggered	Counter	tenant_id, notification_type	Notification fanouts triggered
granit.notifications.deliveries.succeeded	Counter	tenant_id, channel	Deliveries succeeded
granit.notifications.deliveries.failed	Counter	tenant_id, channel	Deliveries failed
granit.notifications.delivery.duration	Histogram (s)	tenant_id, channel, status	Delivery duration

Privacy

Meter: Granit.Privacy

Metric	Type	Tags	Description
granit.privacy.export.requests	Counter	tenant_id	GDPR export requests initiated
granit.privacy.export.fragments.received	Counter	tenant_id, provider	Data fragments received during export
granit.privacy.deletion.requests	Counter	tenant_id	GDPR deletion requests initiated
granit.privacy.export.duration	Histogram (s)	tenant_id, status	Export saga duration

RateLimiting and Bulkhead

Meter: Granit.RateLimiting

Metric	Type	Tags	Description
granit.rate_limiting.requests.allowed	Counter	tenant_id, policy	Requests allowed through rate limiter
granit.rate_limiting.requests.rejected	Counter	tenant_id, policy	Requests rejected by rate limiter

Meter: Granit.Http.Bulkhead

Metric	Type	Tags	Description
granit.http.bulkhead.leases.active	UpDownCounter	tenant_id	Currently active bulkhead leases
granit.http.bulkhead.requests.rejected	Counter	tenant_id	Requests rejected by bulkhead

Vault

Meter: Granit.Vault

Metric	Type	Tags	Description
granit.vault.operations.completed	Counter	tenant_id, operation, provider, status	Vault operations completed
granit.vault.operations.errors	Counter	tenant_id, operation, provider	Vault operation errors
granit.vault.operation.duration	Histogram (s)	tenant_id, operation, provider	Vault operation duration
granit.vault.rotations.detected	Counter	provider	Key rotations detected

Webhooks

Meter: Granit.Webhooks

Metric	Type	Tags	Description
granit.webhooks.fanout.triggered	Counter	tenant_id, event_type	Webhook fanouts triggered
granit.webhooks.deliveries.succeeded	Counter	tenant_id, event_type	Webhook deliveries succeeded
granit.webhooks.deliveries.failed	Counter	tenant_id, event_type, status	Webhook deliveries failed
granit.webhooks.subscriptions.suspended	Counter	tenant_id	Subscriptions auto-suspended
granit.webhooks.delivery.duration	Histogram (s)	tenant_id, event_type, status	Webhook delivery duration

Workflow

Meter: Granit.Workflow

Metric	Type	Tags	Description
granit.workflow.transitions.completed	Counter	tenant_id, outcome, from_state, to_state	Workflow transitions completed
granit.workflow.transition.duration	Histogram (s)	tenant_id, outcome	Transition evaluation duration

Grafana dashboard queries (PromQL)

Service overview panel

# Request throughput by module (last 5 min)
sum by (job) (
  rate(granit_ai_requests_completed_total[5m])
  or rate(granit_notifications_fanout_triggered_total[5m])
  or rate(granit_webhooks_fanout_triggered_total[5m])
  or rate(granit_eventbus_events_published_total[5m])
)

Error rate

# Webhook delivery error rate per tenant
sum by (tenant_id) (rate(granit_webhooks_deliveries_failed_total[5m]))
/
(
  sum by (tenant_id) (rate(granit_webhooks_deliveries_succeeded_total[5m]))
  + sum by (tenant_id) (rate(granit_webhooks_deliveries_failed_total[5m]))
)

Latency percentiles

# P95 vault operation duration
histogram_quantile(0.95, sum by (le, provider) (
  rate(granit_vault_operation_duration_seconds_bucket[5m])
))

AI token consumption

# Token usage per model (last 1 hour)
sum by (model) (
  increase(granit_ai_tokens_input_total[1h])
  + increase(granit_ai_tokens_output_total[1h])
)

GDPR compliance

# Export requests per tenant (last 24h)
sum by (tenant_id) (increase(granit_privacy_export_requests_total[24h]))

# Average export duration
histogram_quantile(0.5, sum by (le) (
  rate(granit_privacy_export_duration_seconds_bucket[24h])
))

Recommended alert rules

Alert	PromQL condition	Severity
High webhook failure rate	rate(granit_webhooks_deliveries_failed_total[5m]) > 0.1	Warning
Vault operation errors	increase(granit_vault_operations_errors_total[5m]) > 0	Critical
Slow AI responses	histogram_quantile(0.95, rate(granit_ai_request_duration_seconds_bucket[5m])) > 30	Warning
Background job failures	rate(granit_backgroundjobs_executions_completed_total{status="error"}[5m]) > 0	Warning
Notification delivery failures	rate(granit_notifications_deliveries_failed_total[5m]) > 0.05	Warning
Webhook subscription suspended	increase(granit_webhooks_subscriptions_suspended_total[1h]) > 0	Warning

See also

• Observability setup — Serilog, OpenTelemetry, ActivitySource auto-registration
• Operations: Observability — Grafana LGTM stack, Collector configuration, alerting
• ADR-001: Serilog + OpenTelemetry — Why this stack was chosen
• Coding standards: Metrics — Developer conventions