Overview

Once spans exist, an exporter sends them to Azure. With Application Insights, telemetry lands in pre-created tables in your Log Analytics workspace — you do not create AppRequests or AppDependencies yourself.

Exam tips

Prefer connection string over instrumentation key alone
Incoming server spans → AppRequests; outbound calls → AppDependencies
Logs → AppTraces; exceptions → AppExceptions; metrics → AppMetrics
Tables are created by Azure when you provision App Insights / workspace
Span attributes appear under customDimensions — no new table required
Classic table names (requests, dependencies) still appear in older docs — same data as App* views

When data lands where

Telemetry	Table (workspace-based)	Classic alias
Incoming requests	`AppRequests`	`requests`
Outgoing calls (HTTP, DB, queue)	`AppDependencies`	`dependencies`
Log messages	`AppTraces`	`traces`
Exceptions	`AppExceptions`	`exceptions`
Metrics	`AppMetrics`	—
Perf counters	`AppPerformanceCounters`	—
Availability tests	`AppAvailabilityResults`	—
Browser timings / page views	`AppBrowserTimings` / `AppPageViews`	`pageViews`

Azure Monitor exporter

The Azure Monitor Trace Exporter (or the distro that wraps it) batches spans and sends them to Application Insights ingestion. BatchSpanProcessor handles batching, retries, and buffering on transient failure.

Configure with a connection string:

Diagram

InstrumentationKey=xxxx;IngestionEndpoint=https://<region>.in.applicationinsights.azure.com/

Python

from azure.monitor.opentelemetry import configure_azure_monitor
import os

configure_azure_monitor(
    connection_string=os.environ["APPLICATIONINSIGHTS_CONNECTION_STRING"],
)

Are App* tables pre-created?

Yes. When you create an Application Insights resource backed by a Log Analytics workspace, Azure provisions a fixed schema of tables. Your exporter only writes rows — there is no CREATE TABLE step for standard App Insights telemetry.

What you can create:

You create	You do not create
Custom log tables (`*_CL`) via Data Collection / ingestion APIs	`AppRequests`, `AppDependencies`, `AppTraces`, …
Custom dimensions via `span.set_attribute`	New columns as physical schema for every attribute
Saved queries / functions (reusable KQL)	The base App Insights schema

Confirm in the portal: Application Insights → Logs — the left schema pane lists tables even before you write a query.

Custom dimensions

Custom dimensions are dynamic properties on telemetry rows. OTel span attributes map here and are queried as customDimensions["key"] or customDimensions.key in KQL.

Python

span.set_attribute("gen_ai.model", deployment_name)
span.set_attribute("rag.top_k", 5)

Kql

AppRequests
| where TimeGenerated > ago(1h)
| extend model = tostring(customDimensions["gen_ai.model"])
| summarize count() by model

Service naming

Without a clear service.name, every app shows as unknown_service. Set it via:

Azure CLI

export OTEL_SERVICE_NAME="embedding-worker"

Or Resource configuration in code. This is critical when multiple AI microservices share one App Insights resource.

Provision resources (CLI)