A three-stage Clos (leaf–spine) data center fabric separates roles to keep forwarding predictable and scalable. The leaf layer is the attachment point for endpoints, so servers connect to leaf devices. This is where the fabric accepts workload traffic, applies edge policies, and provides tenant/service constructs (for EVPN-VXLAN fabrics, the leafs typically act as VTEPs and host IRB gateways as required). The spine layer provides the non-blocking transit core between leafs using L3 underlay routing and ECMP, but it is not the place where servers attach directly.

For north-south connectivity, a typical three-stage Apstra data center architecture includes border leaf switches. These border devices provide the controlled connection point between the fabric and external networks (Internet/WAN/DCI or upstream services). As a result, fabric traffic enters and exits the border devices, not the spines. This design keeps the spines dedicated to high-speed east-west transit and simplifies operations: external routing policies, security controls, and inter-domain handoffs are concentrated at the border, while the spines remain a uniform L3 transit tier. In Junos v24.4 EVPN-VXLAN deployments, this separation also helps maintain clean overlay boundaries and consistent underlay behavior across the fabric.

Verified Juniper sources (URLs):

https://www.juniper.net/documentation/us/en/software/juniper-data-center-assurance/user-guide/topics/concept/dc-network-topology-on-dc-assurance.html

https://www.juniper.net/documentation/us/en/software/jvd/jvd-3-stage-datacenterdesign-with-juniper-apstra/solution_architecture.html

https://www.juniper.net/content/dam/www/assets/white-papers/us/en/design-considerations-for-spine-and-leaf-ip-fabrics.pdf

Apstra 5.1 provides multiple reference designs that define how intent is modeled and how configuration is produced. In the Freeform reference design, the network designer is responsible for creating and validating the device configurations using constructs such as device contexts, property sets, and config templates. Because configuration is authored rather than generated from a fixed data-center abstraction set, Freeform inherently requires practical knowledge of the device’s configuration model and operational behavior—on Juniper platforms, that means being comfortable with Junos-style configuration and show/verification workflows. This makes statement A correct.

Freeform in Apstra 5.1 is also limited to Juniper devices (Junos and Junos Evolved families). In other words, Freeform blueprints do not provide multi-vendor device onboarding and deployment in that reference design at this software level. That makes statement D correct as written in the question’s choices.

By contrast, the data center reference design in Apstra 5.1 is not limited to Junos-only device families; Apstra qualifies multiple NOS families for data center operation via device profiles and packages, and it renders intent accordingly. Therefore statement B is not correct. Finally, while CLI knowledge is always useful for troubleshooting, the data center reference design does not fundamentally require an operator to hand-author device CLI to deploy a standards-based EVPN-VXLAN fabric; that’s one of the core benefits of the intent-based data center reference design.

Verified Juniper sources (URLs):

https://www.juniper.net/documentation/us/en/software/apstra5.1/apstra-user-guide/topics/concept/freeform.html

https://www.juniper.net/documentation/us/en/software/apstra4.2/apstra-freeform-4.2.1/apstra-freeform-guide/topics/concept/freeform-overview-and-design.html

https://www.juniper.net/documentation/us/en/software/apstra5.1/apstra-user-guide/topics/task/device-profile-import-freeform.html

https://www.juniper.net/documentation/us/en/software/apstra5.1/apstra-user-guide/topics/topic-map/devices-qualified.html

In Apstra, tags are an intent-level metadata mechanism used to classify objects and drive automation and reuse. Within a data center blueprint, Apstra supports tagging multiple blueprint objects so operators can apply configuration or policy logic conditionally (for example, applying a connectivity template or a configlet based on a tag match). In this scenario, three valid taggable objects are virtual networks, interfaces, and generic systems.

Virtual network tagging is supported directly from the blueprint’s virtual network table, enabling you to label virtual networks (such as “finance,” “pci,” or “dev”) and then reference those tags elsewhere in blueprint operations and policy application. Interface tagging is also explicitly supported in the blueprint, allowing you to assign tags to switch interfaces and use those tags to control how templates, assignments, or other intent-driven operations apply to those ports. Finally, generic systems (which are modeled endpoint systems such as servers or external routers represented as “systems” in the blueprint) can be tagged so that downstream intent logic can distinguish system roles and apply the correct operations consistently across expansions and changes.

By contrast, property sets are structured data objects used for variable substitution and probe/configlet parameterization, not a primary target for operational tagging in the blueprint UI; and device profiles are catalog artifacts describing hardware/NOS compatibility rather than blueprint objects typically tagged for intent application.

Verified Juniper sources (URLs):

https://www.juniper.net/documentation/us/en/software/apstra5.1/apstra-user-guide/topics/topic-map/tag-interface-add-remove-datacenter.html

https://www.juniper.net/documentation/us/en/software/apstra5.0/apstra-user-guide/topics/task/tag-virtual-network-update.html

https://www.juniper.net/documentation/us/en/software/apstra5.0/apstra-user-guide/topics/topic-map/tag-system-add-remove-freeform.html

In Apstra 5.1, the top-level blueprint banner uses tab indicators (colored badges) to summarize blueprint status across areas such as Staged, Uncommitted, Active, and Analytics. The presence of an Uncommitted indicator signifies that there are staged modifications that have not yet been committed and therefore are not part of the active, deployed intent. That directly corresponds to the statement that changes exist which are not active on the fabric.

At the same time, the banner shows an Active indicator in an alarm state, which reflects that the running fabric has issues requiring attention—commonly surfaced as anomalies (for example, configuration deviation, interface/link faults, protocol/session issues, or service-impacting conditions). In Apstra’s operational model, these issues appear as anomalies that operators should investigate and remediate to restore compliance and health. Therefore, the statement that there are anomalies that must be addressed is also correct.

The remaining options are not implied by this banner alone. Device profile assignment and resource assignment are build-time tasks, but their absence is not what the Uncommitted/Active alert indicators are specifically communicating here. The banner is highlighting uncommitted intent changes and active anomalies that affect the deployed blueprint state and assurance posture.

Verified Juniper sources (URLs):

https://www.juniper.net/documentation/us/en/software/apstra5.1/apstra-user-guide/topics/concept/uncommitted.html

https://www.juniper.net/documentation/us/en/software/apstra5.0/apstra-user-guide/topics/topic-map/anomalies-service-active.html

https://cloudlabs.apstra.com/labguide/Cloudlabs/6.0.0/test-drive-guide/lab1-junos-5_blueprints_.html

In Apstra deployments that use on-box device agents, the agent package installs multiple processes inside the switch’s NOS namespace to provide an isolated runtime environment for Apstra control and telemetry collection. Two of those processes are the Telemetry Agent and the Deployment Agent. The Telemetry Agent is responsible for collecting operational information from the device—such as LLDP neighbor details, routing-related state, and interface information—and sending that telemetry upstream to Apstra. This telemetry is a key input for closed-loop assurance in EVPN-VXLAN fabrics, where Apstra correlates underlay health (interfaces, neighbors, sessions) with overlay services.

The Deployment Agent is responsible for receiving configuration content pushed from Apstra and applying it on the device. In a Junos v24.4 fabric, this is the component that enables Apstra to converge device configuration to the blueprint’s intent (for example, BGP underlay, EVPN signaling, and VXLAN constructs) without requiring manual CLI workflows. Both agents are typically idle most of the time, becoming active when Apstra needs to apply configuration changes or when significant state changes trigger telemetry updates.

Other listed options—“routing agent” and “authentication agent”—are not the named Apstra device-agent processes described for the on-box agent package in Juniper documentation.

Verified Juniper sources (URLs):

https://www.juniper.net/documentation/us/en/software/apstra4.2/apstra-server-and-security-guide/topics/concept/apstra-device-agents.html

Apstra manages devices using IP connectivity over the management network, which is a Layer 3 relationship. Whether you are using on-box agents or off-box agents, the controller (or cluster) communicates with the fabric devices using IP reachability (for example, to exchange management traffic, retrieve discovery state, collect telemetry, and push configuration). This is why Layer 2 adjacency is not required between the Apstra server and the managed switches; the essential requirement is routable IP connectivity and appropriate access (credentials/agent connectivity) to the device management interfaces.

From a platform perspective, Apstra does not need a dedicated physical NIC per managed device. Instead, the server/VM requires connectivity to the management network through a single network adapter, and that interface can route to all managed devices. In a typical data center deployment, the Apstra controller VM sits on a management VLAN/subnet and reaches the entire fabric through routed management. This scales operationally: adding devices does not require adding additional server NICs; it only requires IP reachability and capacity planning for telemetry and agent workloads. Thus, the correct statements are that Apstra uses Layer 3 to communicate with managed devices and that it requires a single network adapter for that management connectivity model.

In Juniper Apstra 5.1, a template is a design abstraction used to create a blueprint. It captures the intended topology shape and design rules without tying the design to a specific vendor’s CLI. Apstra supports multiple template types to match common data center fabric architectures.

A rack-based template is used for the standard three-stage Clos (leaf–spine) approach. In this model, you define the spine logical devices and one or more rack types (containing leaf devices and optional endpoint constructs). This is the dominant pattern for EVPN-VXLAN IP fabrics: leaf switches provide server attachment, VXLAN encapsulation (VTEP function), and optional IRB gateways, while spines provide high-capacity L3 transit with ECMP.

A collapsed template is used for a spine-less (spineless) topology. Instead of a separate spine tier, a collapsed design models a fabric where leaf nodes interconnect in a mesh-like arrangement (as supported by the template type) to provide underlay reachability and redundancy. This can be useful for smaller environments or edge data centers where a full spine tier is unnecessary.

“Compressed” and “device-based” are not Apstra template types. Junos v24.4 is relevant when the blueprint is instantiated and deployed, but the template type selection is an Apstra design-time decision that determines the fabric topology class.