Case study

Tailwind Traders — secrets, identity, and observability

Tailwind Traders operates a multi-tenant SaaS platform across App Service, Azure Functions, and AKS.

Secrets are currently a single Key Vault shared by every application and environment, and each service authenticates with a client secret stored in app settings. A recent near-miss saw a leaked secret grant access to production storage. Security has mandated that no application store a credential of any kind, and that a compromise of one workload must not expose another's secrets.

The database password must rotate every 60 days with zero downtime. An earlier attempt caused an outage because applications cached the old value and kept presenting a revoked credential.

On the observability side, one Application Insights resource is shared by all services, and during incidents the team cannot tell which service is failing or follow a request across service boundaries. Telemetry ingestion cost has also tripled year on year, and a daily cap was applied that silently created gaps in the data.

Requirements

No stored credentials in any application
A compromised workload must not reach another workload's secrets
60-day password rotation with no failed authentications
Attribute telemetry to a specific service and follow a request end to end
Reduce ingestion cost without losing representative data or creating gaps

Q45 Monitor Domain 4 · 20–25% of exam Priority · Real Exam Choose 2

During incidents the team cannot tell which service is failing or follow a request across boundaries. Which TWO changes fix this?

A Set cloud_RoleName per service so telemetry is attributable and the Application Map builds (correct answer)
B Rely on distributed tracing so operation_Id correlates every span of one request (correct answer)
C Create a separate Application Insights resource per developer
D Increase the retention period
E Disable sampling entirely

Shortcut: cloud_RoleName identifies the emitting component and is what drives the Application Map. operation_Id ties every request and dependency in one end-to-end operation together, so a request can be followed across services. Retention governs how long data is queryable, not whether it is attributable.

During incidents the team cannot tell which service is failing or follow a request across boundaries. Which TWO changes fix this?

More Monitor questions