Microsoft AZ-700: Thinking Through Scenarios

An AZ-700 scenario often gives you several network configurations that would all be valid somewhere in Azure. The correct design depends on the particular workload, security boundary and failure condition in the question. This guide works through concrete examples of how to reject plausible but irrelevant choices and prove that the selected fix restores the intended path. It complements a separate review of weak networking skills by making the decision process explicit rather than retelling the topic inventory.

For each scenario, identify five facts before choosing a service: the client and destination, protocol, intended public or private exposure, relevant failure domain, and the observation that would confirm success. These examples are original learning exercises, not actual Microsoft exam questions. Microsoft’s July 2026 AZ-700 exam objectives remain the authority for what candidates should study.

Scenario 1 — Private storage works in Azure but fails from the office

An application VM in VNet-A can read blobs from a storage account using a private endpoint. The same application deployed on an office server times out when using the ordinary storage account hostname. The administrator checks the private endpoint and sees that it is approved. Someone proposes opening public network access on the storage account. Another proposes increasing the office server’s Azure role permissions.

Neither proposal follows from the observation. First inspect the storage hostname resolution from the actual office server. Does it resolve to the private address reachable across the intended VPN or ExpressRoute path, or to a public address that the storage firewall blocks? Check the DNS forwarding and private-zone resolution chain, then verify the route from the office network to the returned private address. Finally confirm TCP connectivity and the calling identity’s data-plane authorization. A successful private endpoint connection from VNet-A proves nothing about the office DNS path.

Decision test: Restore the intended private DNS and network route if they are incorrect, then demonstrate a permitted read from the office and a denied attempt from an unapproved client. Do not make the storage account public to compensate for a name-resolution problem.

Scenario 2 — Two peered networks cannot reach a third workload

VNet-A is peered with a hub, and VNet-B is also peered with that hub. A client in A cannot communicate with a service in B. Both peerings show as connected. The team assumes the second peering should make the VNets mutually reachable. Another engineer proposes enlarging the subnet on the client VM.

Peering is not a general guarantee of transit connectivity. Establish whether the architecture calls for direct peering, approved forwarding through a network virtual appliance, or another supported routing pattern. Inspect effective routes on both sides and compare any user-defined routes with actual appliance capabilities. If the traffic passes through a firewall, check its policy and the return path. Changing a subnet size without evidence addresses neither transit behavior nor a missing route.

Decision test: Make a path diagram with next hops in both directions. Choose the design that maintains the intended inspection boundary, then confirm packets reach the service and responses return to the source. A one-way capture is not a successful connection.

Scenario 3 — Application Gateway reports an unhealthy backend

A regional HTTPS application works when a developer calls a backend VM directly but returns errors through Application Gateway. The VM is running and its NSG allows the gateway subnet’s traffic. One suggestion is to replace the gateway with Azure Load Balancer, because the application is ‘only web traffic.’ Another is to disable the WAF entirely.

Begin with the configured Application Gateway backend health and its probe. Does the probe use the host header and path the application expects? Does the backend certificate match the configured hostname when HTTPS is used to reach it? Is DNS resolving the backend address from the gateway, and does the backend accept traffic on the chosen port? A service can return a valid homepage to a human and fail a probe aimed at a protected path. WAF inspection of client requests is a separate boundary from gateway-to-backend health.

Decision test: Correct the probe, hostname, trust or backend connectivity based on the observed failure. Prove the gateway can select a healthy backend and complete an application request. Replacing the product or disabling security without isolating the fault creates more uncertainty.

Scenario 4 — Regional failover appears to work but users keep failing

A company uses Azure Traffic Manager to direct clients to two regional endpoints. During a test, endpoint health changes and new DNS answers point to the secondary region, but some users continue contacting the failed primary. An architect proposes moving all traffic behind a regional load balancer. A support engineer assumes Traffic Manager is malfunctioning because users did not migrate instantly.

Traffic Manager makes a DNS decision and is not in the application data path. Recursive resolvers and clients can cache earlier answers, and existing sessions may outlast a health-check decision. A regional load balancer does not replace a global DNS failover requirement. Decide whether the workload needs DNS-based routing for the supported protocol or an HTTP-aware global edge proxy such as Front Door, considering its own behavior and cost. Neither product can fix an unavailable database in the secondary region.

Decision test: Measure DNS caching and new-connection behavior separately from existing sessions, test the secondary application’s business dependencies, and define a realistic recovery target. A green endpoint-health signal is not proof that all users have recovered.

Scenario 5 — The VPN is connected, but an internal API times out

A site-to-site VPN reports connected. An office monitoring host can reach one Azure subnet but cannot call a private API behind a network appliance in a second subnet. The team’s first suggestion is a larger VPN gateway SKU. The incident began immediately after a route-table change. That timing is evidence worth pursuing, not final proof.

Determine the source and destination prefixes, the VPN’s learned routes, the effective route on the target subnet and the next hop expected for the return packet. Compare those values with the route-table configuration before the change. Confirm that the firewall or appliance is receiving the flow and that its stateful rules permit the expected return. A larger gateway would not repair a route pointing at the wrong appliance or an asymmetric path.

Decision test: Restore the smallest supported routing change that recovers the bidirectional path, then prove both the API request and response succeed while unintended network paths stay denied. Record the relevant route decision and client result in the incident note.

Scenario 6 — Private endpoint, service endpoint and public IP all seem plausible

A regulated application must access a supported platform service over a private address from selected VNets. The design options include a service endpoint, a private endpoint and a public endpoint restricted by IP rules. Each can support legitimate security designs, but they do not create the same network boundary. A service endpoint typically connects an authorized VNet subnet to a public service endpoint through supported service routing; a private endpoint gives an eligible service a private IP in a VNet and requires appropriate DNS. An IP-filtered public endpoint still has a different exposure model.

When the requirement explicitly calls for a private IP and private DNS-controlled path, investigate a private endpoint with an approved DNS and routing design. When the application instead requires subnet-based authorization to a supported public service endpoint, a service endpoint might meet that narrower design. Review whether a proxy, an on-premises client or a different tenant also needs connectivity before deciding. Selecting the newest feature by name is not a substitute for reading the access requirement.

Decision test: From authorized and unauthorized test clients, inspect name resolution, route, connection and data permissions. Show where the boundary is enforced. Keep the data-plane identity requirement separate from network eligibility.

Scenario 7 — A connection fails after tightening a network policy

A maintenance window introduces a stricter NSG rule. An application begins timing out during a database call. A colleague immediately wants to allow all outbound traffic. Instead, identify the required source, destination, protocol and port. Check effective NSG rules, the relevant route and any centralized firewall policy. Determine whether the connection is blocked at the source, at an intermediate hop or near the destination. A policy change can expose a previously undocumented dependency, but the correct response is to document that dependency and authorize only its necessary path.

Decision test: Add a narrowly scoped rule only when the business requirement and packet evidence justify it. Verify the database operation succeeds and a deliberately unauthorized destination is still blocked. A firewall rule that makes every possible future connection work is not a successful least-privilege design.

Score your reasoning, not just your final selection

After each exercise, write the requirement, competing plausible choices, decisive observation, minimal correction and verification evidence. A robust answer distinguishes what can be inferred from a green dashboard from what the user experiences. It should also state what new fact would change the recommendation. For example, a requirement for non-HTTP global routing can change the Front Door versus DNS selection; a hard requirement for private service addressing changes the service-endpoint discussion.

These scenarios belong within Microsoft’s current AZ-700 areas for core infrastructure, connectivity, application delivery, private access and network security. Repeat one exercise with a changed constraint rather than memorizing a single correct service. The aim is a defensible network decision with observable results—not a list of Azure products dressed up as exam preparation.

img