Overview

Distributed AI systems fail in pieces — the API returns 500 because a dependency timed out three hops away. Trace data lets you reconstruct the whole path using OperationId (the Azure name for the OTel trace ID) and find the break point.

Exam tips

Start from Failures / Performance end-to-end transaction view, or from KQL on AppDependencies
Correlate with OperationId across AppRequests, AppDependencies, AppExceptions, AppTraces
Parent succeeds + child fails → look at the failing dependency
Duration spike without exception → latency bottleneck, not necessarily an error
Broken traces at a boundary → missing traceparent propagation or uninstrumented service

Debugging workflow

Portal waterfall — Application Insights → Performance or Failures → open a request → End-to-end transaction
Or hunt by pattern in KQL — e.g. failing dependency target
Pivot on OperationId — union all telemetry for that operation
Find the break — failing child, ballooning duration, or missing spans after a hop
Fix instrumentation if traces stop cold at a queue or service with no children

Operation ID

OperationId (classic: operation_Id) is the shared correlation ID for one end-to-end request across Application Insights tables. It maps to the OpenTelemetry trace ID.

Kql

AppDependencies
| where Success == false and Target == "payments-api"
| project OperationId, TimeGenerated, Name, ResultCode

Then reconstruct the full flow:

Kql

union AppRequests, AppDependencies, AppExceptions, AppTraces
| where OperationId in ("id1", "id2", "id3")
| order by OperationId, TimeGenerated asc

Practical KQL patterns

Exceptions joined to the triggering request

Kql

AppExceptions
| where TimeGenerated > ago(6h)
| join kind=inner (AppRequests | where TimeGenerated > ago(6h)) on OperationId
| project TimeGenerated, ProblemId, OuterMessage, Name, ResultCode

P95 latency by operation (spot regressions)

Kql

AppRequests
| where TimeGenerated > ago(1d)
| summarize p95 = percentile(DurationMs, 95) by Name
| order by p95 desc

Full trace for one operation

Kql

union AppRequests, AppDependencies, AppTraces, AppExceptions
| where OperationId == "abc123def456"
| project TimeGenerated, ItemType = type, Name, Message, DurationMs
| order by TimeGenerated asc

union stacks rows from multiple tables. The special type column tells you which table each row came from — useful for a unified timeline.

Common failure modes

Symptom	Likely cause
Trace ends at Service A; nothing in Service B	Missing `traceparent` or B not instrumented
Request OK, dependency failed	Downstream error — inspect `AppDependencies` / `AppExceptions`
High duration, Success true	Slow path (DB, model, cold start) — not a hard failure
Many `unknown_service` rows	Missing `OTEL_SERVICE_NAME` / `service.name`
Gaps in high-traffic apps	Aggressive sampling — increase rate temporarily while debugging

Debug distributed flows with trace data