Application Insights

Every Microtec environment has a dedicated Application Insights workspace that collects telemetry from all .NET backend services and Angular frontends. Application Insights is integrated with OpenTelemetry for backend services and with the Angular SDK for frontend correlation.

---

Per-Environment Workspaces

Environment	Resource Name	Resource Group
dev	mic-erp-be-dev-ai	mic-erp-be-dev-monitoring-rg
stage	mic-erp-be-stage-ai	mic-erp-be-stage-monitoring-rg
preprod	mic-erp-be-preprod-ai	mic-erp-be-preprod-monitoring-rg
uat	mic-erp-be-uat-ai	mic-erp-be-uat-monitoring-rg
production	mic-erp-be-prod-ai	mic-erp-be-prod-monitoring-rg

Each workspace is backed by a Log Analytics workspace with 90-day retention (dev/stage) or 180-day retention (prod/uat/preprod).

---

Connection String Injection

Application Insights connection strings are not hardcoded. They are stored in Key Vault and injected via Container App environment variable references:

KV Secret name: ApplicationInsights--ConnectionString
CAE env var:    ApplicationInsights__ConnectionString

Connection string format:

InstrumentationKey=<guid>;IngestionEndpoint=https://uksouth-1.in.applicationinsights.azure.com/;LiveEndpoint=https://uksouth.livediagnostics.monitor.azure.com/

Connection String vs Instrumentation Key

Always use the connection string (not the deprecated instrumentation key alone). The connection string includes the regional ingestion endpoint, which ensures data is routed to the correct region (UK South for Microtec) without latency. Using only the ikey routes through the global endpoint which adds ~50ms on average.

---

.NET Integration

Application Insights is wired up via OpenTelemetry in Program.cs using the Microtec.Web.Core shared package:

// In Microtec.Web.Core / ServiceCollectionExtensions.cs
builder.Services.AddOpenTelemetry()
    .WithTracing(tracing => tracing
        .AddAspNetCoreInstrumentation()
        .AddHttpClientInstrumentation()
        .AddEntityFrameworkCoreInstrumentation()
        .AddAzureMonitorTraceExporter(opts =>
        {
            opts.ConnectionString = builder.Configuration
                ["ApplicationInsights:ConnectionString"]!;
        }))
    .WithMetrics(metrics => metrics
        .AddAspNetCoreInstrumentation()
        .AddRuntimeInstrumentation()
        .AddAzureMonitorMetricExporter(opts =>
        {
            opts.ConnectionString = builder.Configuration
                ["ApplicationInsights:ConnectionString"]!;
        }));

The shared package handles this automatically. Individual services do not configure Application Insights directly.

---

Angular Frontend Integration

The Angular applications use @microsoft/applicationinsights-web for browser-side telemetry:

// libs/shared-lib/src/lib/services/telemetry.service.ts
import { ApplicationInsights } from '@microsoft/applicationinsights-web';

@Injectable({ providedIn: 'root' })
export class TelemetryService {
  private appInsights: ApplicationInsights;

  constructor(private config: ConfigService) {
    this.appInsights = new ApplicationInsights({
      config: {
        connectionString: config.get('applicationInsightsConnectionString'),
        enableAutoRouteTracking: true,
        enableCorsCorrelation: true,
        correlationHeaderExcludedDomains: ['auth.microtecstage.com']
      }
    });
    this.appInsights.loadAppInsights();
  }
}

Frontend telemetry uses a separate connection string pointing to the same Application Insights resource but with a different role name (Angular) to distinguish frontend from backend telemetry.

---

Telemetry Types Collected

Type	Backend	Frontend
Requests	All HTTP requests with duration, status	Page views, route changes
Dependencies	SQL queries, Redis calls, HTTP calls to other services	XHR/fetch calls to API
Exceptions	Unhandled exceptions with stack traces	JS errors and unhandled promise rejections
Custom metrics	Business metrics (tenants active, invoices processed/hour)	User action metrics
Traces	Structured log forwarding from Serilog	Console.error forwarding
Availability	Health check pings (via Availability Tests)	—

---

Custom Dashboards

Three shared dashboards are pinned to the Azure Portal:

1. Operations Overview Dashboard

Available at Dashboards → Microtec ERP - Operations:

Tile	Metric
Request rate	Requests/min across all services
Failure rate	5xx responses as % of total
P95 latency	95th percentile response time per service
Active exceptions	Exceptions in last hour by service
Dependency health	SQL / Redis / ASB call success rate

2. Service Health Dashboard

Per-service latency, error rate, and dependency breakdown. Drill-down from Operations Overview.

3. Business Metrics Dashboard

Custom events tracked by the application:

• Tenants active in last 24 hours
• Invoices created per hour
• Zatca submission success/failure rate
• User login success/failure rate per realm

---

Availability Tests

Standard availability tests ping critical endpoints every 5 minutes from multiple Azure regions:

Test Name	URL	Alert threshold
Gateway Health	https://gateway.{domain}/health	2 consecutive failures
Keycloak Ready	https://auth.{domain}/health/ready	2 consecutive failures
Frontend	https://erp.{domain}/index.html	3 consecutive failures

Failures trigger alerts via the alerting action groups.

---

Sampling

To control costs, Application Insights uses adaptive sampling in all environments:

Environment	Sampling rate
dev	100% (no sampling)
stage	100% (no sampling)
preprod	50%
uat	100% (for UAT accuracy)
production	Adaptive (10–100% based on volume)

Production sampling is configured to always include:

• All failed requests (HTTP 5xx)
• All requests with exceptions
• All requests exceeding P99 latency threshold (>5s)

// Sampling is configured in Microtec.Web.Core
options.SamplingSettings = new SamplingSettings
{
    IsAdaptiveSamplingEnabled = true,
    MaxTelemetryItemsPerSecond = 20
};