instrumenting-azure-app-insights

Instrumenting Azure Application Insights

Workspace-based Application Insights with cost guardrails. Classic (non-workspace) App Insights is retiring, so this is the only forward-compatible setup.

When to invoke

• Adding observability to a new project
• Migrating off classic Application Insights
• Setting up failure-rate alerts that page an owner
• Auditing observability cost on an existing app

The two-resource setup

App Insights now sits on top of a Log Analytics workspace:

Log Analytics Workspace  ← cost guardrail lives here (dailyQuotaGb)
       ↑
       │  WorkspaceResourceId
       │
Application Insights component  ← what the SDK talks to
       ↑
       │  APPLICATIONINSIGHTS_CONNECTION_STRING
       │
Your app's instrumented SDK

Workspace owns:

• Retention (retentionInDays: 30 default; max 730)
• Daily quota cap (workspaceCapping.dailyQuotaGb: 1)
• All actual data storage

App Insights component owns:

• The connection string the SDK uses
• Per-app filtering / sampling rules

Cost guardrail: dailyQuotaGb

The single most important setting. Without it, a runaway log loop can ingest 100+ GB and bill hundreds of dollars overnight.

resource workspace 'Microsoft.OperationalInsights/workspaces@2023-09-01' = {
  properties: {
    sku: { name: 'PerGB2018' }
    retentionInDays: 30
    workspaceCapping: { dailyQuotaGb: 1 }    // ← hard daily cap
  }
}

Rule of thumb:

• Hobby / low-traffic: 1 GB/day cap (~$2/month at default rates)
• Small SaaS: 5 GB/day cap
• Production with structured logging: start at 5, monitor, raise as needed

Once the cap is hit, ingestion is paused for the rest of the UTC day. You'll lose log lines but not the bill.

Optional metric alerts

When alertEmail is non-empty, the module creates:

1. Action Group — sends to the email (could be extended to SMS, webhook, etc.)
2. 5xx alert — fires if requests/failed > 5 in any 15-min window
3. Exceptions alert — fires if exceptions/server > 5 in any 15-min window