Microsoft AZ-104 Exam Dumps, Practice Test Questions

Always On keeps the application process warm on existing workers; it does not create or preserve a minimum of three workers under autoscale.

Default capacity is not an enforced floor after autoscale actions. If minimum remains one, a scale-in rule may still reduce the app below three workers overnight; set the minimum to three.

The maximum caps scale-out but does not prevent automatic scale-in to one instance. It cannot establish the requested overnight floor.

Apps in one plan share workers by design. Billing follows the plan's resource allocation rather than automatically duplicating it for each hosted app.

Request volume is useful for workload sizing, but dedicated App Service plan compute charges depend primarily on the allocated worker SKU and instance capacity. Two apps sharing a plan do not receive independent per-request worker bills merely because their request counts differ.

A web app hosted on existing dedicated plan workers shares that plan's allocated compute and costs. Its small size does not automatically change the billing model to consumption or create a separately metered hosting plan.

Routing affinity does not copy application memory between workers. It merely influences which instance receives subsequent requests.

ARR affinity encourages requests to reach the same worker, but cannot restore session data when that worker recycles. Durable or external session state is still required.

Worker-local temporary storage may be lost or isolated when instances are recycled or changed. ARR cookies influence routing, not durable replication of state across workers.

Making the first app identity more privileged broadens only that principal’s data authorization. System-assigned identities belong to individual App Service resources; a second app on the same plan does not automatically become the first app’s principal. Configure and authorize an identity for each app that requires access instead of granting unnecessary subscription-wide access.

A user-assigned identity can deliberately be shared across apps, but each app must explicitly attach or reference that identity. The shared hosting plan does not automatically attach it to the second app or grant that app tokens. This proposal configures only the first app and therefore fails the requested second app's data-plane access.

Reader over an App Service resource is an ARM permission, not authority to impersonate that resource's system-assigned managed identity. The second app cannot request tokens as the first app's identity merely by reading its settings. It needs its own authorized managed identity or an explicitly attached shared user-assigned identity and appropriate data access roles.

Deleting staging values can break staging tests and does not provide controlled slot-specific configuration behavior during a swap.

Bundling a production secret with swappable code risks moving it into the staging slot and breaks environment-specific isolation. Slot-sticky configuration is the supported setting.

Using one database for both environments may let staging tests touch production data and contradicts the requirement that each slot retain its own database.

Traffic percentages route selected requests among active deployment slots; they do not perform the two-slot exchange that places the former production content into staging. The described release-and-swap-back workflow requires an actual slot swap.

Slot-sticky settings control which configuration remains bound to a slot during swaps. Marking settings sticky does not move application content or initiate promotion. Validate staging, perform the swap, and swap back if postrelease checks fail.

Auto swap promotes supported staging deployments after warm-up without the manual approval step. It can be useful for a fully automated pipeline but conflicts with the stated validate-THEN-promote decision and deliberate rollback workflow here.

A slot swap exchanges content/configuration according to swap rules for the application, not a single opt-in browser session. Frequent swaps would expose new code broadly and disrupt the release workflow. The team explicitly wants production to stay intact with no automatic beta traffic distribution.

App Service slot routing percentages are not a built-in Entra group-based audience filter. Testers can opt in by using the documented x-ms-routing-name query parameter, which sets slot affinity without changing the global percentage.

A slot swap warm-up ping prepares application instances during swaps; it is not a per-client routing selector. At 0% automatic traffic, use the x-ms-routing-name query parameter/cookie for an opted-in beta session.

Production retains its configured identity, so relying on staging principal permissions creates unpredictable access after content moves.

Custom hostnames and identity/VNet settings are slot-specific and not a fix for production’s missing private resource permissions.

Slot-setting stickiness affects supported application settings; the identity principal itself is not swapped with code or ordinary settings.

An ordinary direct swap can promote the staged version, but it does not provide the requested pause to inspect production-target configuration before completion.

Automatic swap can promote staging changes as part of deployment, but it removes the explicit opportunity to pause and cancel after validation with target settings.

Manual copying can approximate configuration but may mishandle slot-specific settings and lacks the supported preview pause and cancellation workflow.

Slot-specific settings stay with their slot but do not by themselves initiate a swap when a deployment completes.

Swap with preview pauses for validation and still needs an explicit completion, so it does not satisfy hands-off promotion after every code push.

Traffic percentages can route requests to a slot but do not execute an automatic deployment-slot swap. For the stated post-deployment warm-up and promotion, configure Auto Swap on a supported Windows App Service slot.

Slots can have their own accessible endpoints. Requiring production traffic for testing would defeat the isolation of a preproduction slot.

A production-host CNAME routes users to the production app, not automatically to the isolated staging slot's code. QA should use the staging slot's own endpoint or an explicitly configured custom staging-slot hostname before any swap.

App Service abstracts the worker infrastructure; VM naming is not how slot endpoints are addressed. The slot hostname remains the correct target.