Microsoft AZ-104 Exam Dumps, Practice Test Questions

Availability zones are a compute placement/resiliency choice and do not change the Azure Monitor DCR regional location rule. The important distinction is region of the data collection rule versus region of its workspace destination, not where the VM sits within an availability zone.

For the supported Azure Monitor Agent collection scenario, a DCR must be in the region of its destination workspace. When workspaces are in different regions, use a separate DCR associated with the same VM for the West Europe workspace.

An Azure VM can have multiple DCR associations; Azure Monitor Agent processes the rules associated with the machine. Creating another VM would add cost and complexity without satisfying the intended two-workspace monitoring pattern for the existing workload.

AMA can collect performance counters, but those records do not contain all of the process-to-process connections used by the legacy Map experience. That feature depended on separate dependency instrumentation and now has explicit retirement/onboarding restrictions. Seeing guest CPU and memory in Log Analytics is not proof that a newly onboarded VM can use Map.

The retirement announcement targets the Map experience and Dependency Agent, not the AMA-based VM insights performance features. The scenario already receives valid guest performance counters in Log Analytics, which can continue to be collected under supported data collection rules. Treating all VM insights as retired would unnecessarily remove operational telemetry.

A new Dependency Agent installation does not override Microsoft's new-VM portal onboarding restriction or extend the announced 2028 end of support. Microsoft explicitly discourages new installations of the retiring Dependency Agent. A new deployment should choose current supportable dependency visibility rather than relying on a legacy Map workaround.

A configured failure threshold is not a binary Pass-until-Fail switch. Connection Monitor provides an early Warning state when failed checks exceed 80% of the configured threshold. An 18% failure rate with a 20% threshold already exceeds the 16% warning boundary, so reporting Pass would conceal a developing connection-quality problem.

Fail is reached when a configured threshold for failed checks or round-trip time is exceeded. The scenario reports 18% failed checks against a 20% failure threshold, with acceptable round-trip time. A few failed probes do not automatically produce a Fail status; the documented early-warning band applies instead.

Indeterminate denotes missing monitoring data rather than a measured percentage below the Fail limit. This enabled test has delivered an 18% checks-failed value and a valid RTT result. The observation is sufficient to evaluate thresholds, and the defined 80%-of-failure-threshold warning rule leads to Warning.

A TCP check verifies transport-level connection establishment, but a server can accept port 443 connections and still return HTTP 503. TCP success does not assert that the higher-layer application returned an acceptable HTTP response.

ICMP measures network-level reachability and potentially latency, not HTTPS application status. The API server may answer pings while every customer request fails; a valid user-facing health test needs an HTTP response criterion.

A disabled test group does not run its checks and reports Not Running rather than revealing application HTTP success or failure. Suppressing tests conflicts with the need to detect application outages.

Packet capture is a diagnostic collection operation, not a continuously evaluated availability alert with notification recipients. A capture might help investigate a reported failure but does not replace repeated Connection Monitor tests plus an Azure Monitor alert/action group. The team also does not need broad resource-change permissions to observe a TCP path.

Next hop diagnoses the route selected for a specific VM-to-destination flow; it does not provide ongoing TCP endpoint checks or a notification channel by itself. It can supplement troubleshooting after a failed probe, but a Connection Monitor failure signal and configured Azure Monitor alert/action group meet the requirement without changing firewall policy.

Connection Monitor thresholds determine whether its TCP test results show a degraded or failed state. They do not, by themselves, configure email recipients or an incident notification route. The team needs an Azure Monitor alert using that telemetry and an action group for notification, keeping the firewall controls unchanged.

The port controls the target test, not which VM generates the probe. The old source still would not represent the replacement VM.

A test using the retired source resource ID will still report from the wrong source. The test group must point to the replacement VM endpoint to make probe measurements representative.

Agent installation on the replacement VM can be a prerequisite, but it does not reassign the Connection monitor test group's source endpoint. Update that endpoint reference as well.

An overall monitor health figure cannot establish which dependency or specific port failed. Connection Monitor derives results from source/destination/test-configuration combinations. Maintaining individually identifiable tests for the internal API on 443 and partner endpoint on 8443 provides the required separate trends.

A test configuration specifies its transport port; applying the same port-443 configuration to two destinations tests that port for both. It would miss a failure specific to the partner’s 8443 listener. Define configurations/tests with the correct port for each destination and retain their individual measurements.

A successful ICMP probe to a gateway does not establish that either destination application accepts TCP connections, much less distinguish the internal API from the external partner. The required telemetry is destination- and port-specific; create individually identifiable TCP tests using the respective endpoints and service ports.

IP flow verify answers a narrower question about the effective NSG verdict for a hypothetical packet. An allowed NSG flow doesn't prove that the guest firewall, route or database listener accepts the connection. The symptom persists after an NSG permit, so a more end-to-end diagnostic is needed.

VNet flow logs show network-policy flow observations, not proof that the database process is listening or that guest firewall rules allow the session. Connection troubleshoot is better suited to diagnosing the end-to-end route, guest block or destination problem.

A missing SYN-ACK in a source capture shows an unsuccessful handshake but does not alone distinguish an intermediate route drop, a destination guest firewall, and an absent listening service. Connection troubleshoot adds specific network-path diagnostics for the failed source-to-target session.

This reverses the source/destination requirement. The remote partner cannot make the Azure source execute probes by installing an extension at the far end, and supported Azure VM sources do require their monitoring components.

Azure VMs are supported as sources, but the destination list is broader and explicitly includes external URLs, FQDNs and IP addresses. Restricting both endpoints to agent-equipped Azure VMs would incorrectly exclude a documented Connection Monitor use case.

DNS zone peering is not a Connection Monitor substitute and does not measure HTTP reachability or latency to an external HTTPS endpoint. The supplied FQDN can be used directly as the monitored destination.

Increasing metric aggregation granularity changes how the displayed measurements are summarized, not the amount of history that Azure Monitor retains for Connection Monitor metrics. A six-month seasonal analysis requires persisted monitoring logs with a suitable Log Analytics workspace/table retention setting, not just daily chart bins.

Pausing a test group stops or interrupts measurements rather than preserving a complete six-month record. Metrics do not become permanent because a test group is disabled. This approach would create additional data gaps exactly where the enterprise wants continuous long-run reachability evidence.

Increasing test frequency produces more samples and potentially greater monitoring costs, but does not change the default history limit for the metric store. The requested six-month reachability and latency trends need a long-term monitoring-data destination such as Log Analytics and an appropriate retention policy.

Threshold exceedance is associated with a Warning or Fail evaluation when tests are running. A disabled group is not executing checks and cannot supply a meaningful RTT failure measurement regardless of threshold changes.

When connectivity checks run and all fail, the test can produce Fail under the documented threshold logic. Not Running is instead the state for a disabled test group, so changing a destination would be the wrong first response.

Action groups govern alert notifications, not whether Connection Monitor executes a test group. An empty receiver list may affect paging but cannot explain why the test itself is explicitly in the disabled state.