Practice Exams:

Microsoft AZ-104: ExpressRoute Resilience Patterns

Cloud & Architecture

Private connectivity can be reliable and still contain one unexpected point of failure. ExpressRoute resilience depends on physical diversity, independent routing paths and tested failover behavior—not simply the presence of two BGP sessions on a circuit. Redundant sessions across one physical location may still share a provider or facility failure. ExpressRoute resilience should be assessed through independent circuits, peering, routing choices, and a controlled test of the exact failure the design must survive.

On this page
  1. Distinguish link redundancy from location redundancy
  2. Choose circuit diversity according to the consequence of an outage
  3. Design BGP routing for planned and unexpected path changes
  4. Treat the Azure gateway as part of the failure path
  5. Decide whether an encrypted internet VPN is the backup path
  6. Practice both network failover and application recovery
  7. Monitor capacity, changes and hidden dependencies

An ExpressRoute circuit is provisioned with redundant connections into Microsoft’s edge, and it exchanges routing information through supported peerings. That helps protect against certain local port or link failures. It does not guarantee continuity if the same provider building, metropolitan peering location or on-premises router becomes unavailable. Two logical connections that share the same duct or device remain one physical failure domain. A Microsoft AZ-104 ExpressRoute plan should account for circuit, gateway, peering, and on-premises failures separately.

Draw the connectivity path from an application to its Azure destination. Include the enterprise access router, provider handoff, fiber routes, peering locations, ExpressRoute circuits, gateway or virtual hub, and relevant DNS services. Mark which links, power systems, devices and operations teams are shared. Many organizations discover that supposedly independent circuits converge at the same cross-connect or carrier aggregation device.

Classify the risks you need to tolerate: a single port failure, one provider outage, loss of a peering site, gateway maintenance or route misconfiguration. A resilience statement should name those events. “Dual ExpressRoute” alone is too vague to tell an incident commander which failure still takes the service down.

Choose circuit diversity according to the consequence of an outage

For demanding production workloads, Microsoft documents circuit resilience options, including two circuits across separate peering locations and ExpressRoute Metro in supported markets. The current ExpressRoute circuit creation guidance calls its maximum-resiliency option a two-circuit architecture across different ExpressRoute edge locations. That is a meaningful safeguard against location-level disruption, provided the customer and provider paths are independently engineered.

Circuit diversity may require different carriers, separate premises entrances, diverse customer edge routers or geographic separation. These are enterprise architecture choices as much as Azure resource selections. An Azure portal resource ID does not reveal whether both circuits ultimately depend on the same upstream cable. Ask the carrier for evidence of physical diversity and document what is contractual versus merely preferred.

Not every workload warrants the cost and complexity of separate locations. A noncritical environment may tolerate a single circuit plus a measured restoration procedure. A critical transaction system may require separate locations and an additional backup technology. Document cost, achievable recovery objectives and residual dependencies instead of presenting one universal connectivity blueprint.

Design BGP routing for planned and unexpected path changes

BGP advertisements determine which prefixes are reachable over a circuit. A sound topology needs consistent address planning, permitted prefixes, routing filters and explicit expectations for preferred and backup paths. Review customer-edge devices and Microsoft’s documented routing behaviors before manipulating AS-path or other BGP attributes. A policy that works in one direction may not create symmetry in the other.

Failover often exposes assumptions made about equal-cost routing, firewall state or NAT. If the outbound flow shifts to a secondary circuit while the return flow keeps its old path, a stateful inspection device may drop the session. Map both directions for representative applications, especially those traversing central security appliances. Avoid interpreting a successful BGP peer status as proof that a transactional application can complete a request.

Use ExpressRoute recovery guidance to define the network ownership and configuration preservation requirements. Capture circuit identifiers, peering VLANs, provider contacts, authorized prefix lists, route preferences and rollback configurations in version-controlled operational documentation. A connectivity outage is the wrong time to discover that only one engineer knows the route policy.

Treat the Azure gateway as part of the failure path

Multiple physical circuits do not remove the need to consider the ExpressRoute gateway or managed hub that delivers routes into the Azure environment. Review available redundancy, region and zone support, gateway scale and supported configurations. A zone-redundant gateway SKU can reduce particular availability-zone risks, but it does not replace diversity of circuits or the customer’s own WAN equipment.

Bandwidth planning must consider normal and degraded operation. If two routes share the load, the surviving path might suddenly carry most traffic when one disappears. A standby circuit that cannot handle the critical workload during failover is technically present but operationally inadequate. Prioritize traffic according to business impact and verify bandwidth and throughput constraints of gateways as well as provider circuits.

Review connections from each circuit to each needed gateway. Some multi-homed gateway capabilities and resiliency models have changed over time. Follow the current support matrix and do not assume an older design template is automatically supported by the latest Azure feature set or a particular connectivity provider.

Decide whether an encrypted internet VPN is the backup path

A site-to-site VPN can provide an alternate path when ExpressRoute is disrupted, but it uses the public internet and a different availability, performance and security profile. Design it as a defined degraded mode rather than claiming identical service. Establish whether critical prefixes, DNS resolvers, identity systems and cross-premises applications remain reachable when routes change.

In supported gateway topologies, VPN Gateway design guidance documents ExpressRoute and site-to-site coexistence. BGP and route priorities require careful configuration to avoid a backup route accidentally carrying normal traffic or remaining unused when the primary path fails. The backup’s encryption, key rotation, firewall policy and internet uplink should be independently tested.

An alternate VPN often has far less bandwidth than ExpressRoute. Determine which service tiers must remain operational during failover. A deliberate protection policy may allow critical administration, authentication and transactions while suspending large replication jobs. Test actual response times instead of regarding successful tunnel establishment as sufficient.

Practice both network failover and application recovery

A controlled failover drill should remove or isolate a chosen circuit or routing path through an approved method and verify both the routing state and the applications that depend on it. Begin with a limited test window, a rollback owner and an agreed impact budget. Record route withdrawal and convergence times, packet loss, session resets, DNS behavior and recovered business transactions.

Run distinct tests for a single connection, an entire circuit, a peering location scenario where simulation is feasible, and an ExpressRoute-to-VPN backup. Avoid destructive testing against production without change approval. A failure can appear smaller than expected in synthetic pings but much larger in applications using long-lived sessions or firewall state; collect evidence at both layers.

Recheck the failback path. Returning traffic to the preferred circuit can create a second interruption after the initial recovery if BGP policy or session affinity is poorly understood. A resilience design should specify how the preferred route is restored and what confirms the application is stable before the incident is closed.

Monitor capacity, changes and hidden dependencies

Monitor circuit and gateway metrics, BGP state, route stability, traffic levels and provider maintenance advisories. A completely down circuit is easier to detect than a partial impairment with intermittent packet loss or flapping routes. Correlate provider alerts with Azure telemetry and application response metrics to distinguish a WAN failure from a service dependency issue.

Review configuration changes that can defeat redundancy: adding a route filter only on one circuit, changing the customer’s autonomous system policy, renumbering on-premises networks or moving an Azure workload behind another gateway. Every change to a high-availability design should prompt a question about how the second path behaves. A spare path that has silently stopped receiving routes is no longer a spare.

Some designs expand into managed hubs, where Azure Virtual WAN changes the routing control plane and connection model. The same core lesson applies: availability is created by distinct failure paths plus verified recovery behavior. It cannot be inferred from redundant-looking lines on a network diagram.

Continue learning

Related guides

Azure VPN Gateway DesignPlan gateway topology, tunnel routing, and failover tests.Azure Route Server for Hybrid NetworksTest BGP route exchange, appliance peering, and return paths.Azure Virtual WAN Design DecisionsAssess Virtual WAN hubs, route propagation, and segmentation.

Related Posts

• Microsoft AZ-104: What Azure Administrators Actually Do

• Exploring the AZ-900 Certification – Your Gateway to Cloud Computing Mastery

• Cloud-Native App Architecture Using Microsoft Azure Cosmos DB

• Microsoft AZ-104: Azure Monitor Alerts

• Microsoft AZ-104: Azure Backup vs. Site Recovery

• Microsoft AZ-104: NSGs, ASGs, and Azure Firewall

• Microsoft AZ-104: Subscriptions, Policy, and Locks

• Microsoft AZ-900: Shared Responsibility by Cloud Service Model

• Microsoft AZ-104: Azure RBAC for Platform Teams

• Microsoft AZ-104: Azure Route Server for Hybrid Networks