BGP runs over TCP (port 179). Stateful firewalls often have configurable idle timeouts for TCP sessions. If the firewall’s TCP timeout is shorter than BGP’s keepalive/hold timers, the firewall will tear down its state table entry, causing subsequent BGP packets (like keepalives) to be dropped, breaking the session. The peers will then restart the TCP connection, repeating the cycle. This is a classic real-world scenario distinct from simple configuration errors on the routers themselves.

In SR-MPLS, the Segment Routing Header is encoded as a stack of MPLS labels. Like in traditional MPLS, the Bottom-of-Stack (BoS) bit in the MPLS label header is set to ‘1’ on the last label in the stack (the innermost label). This signals to the processing router that after this label is popped, what follows is the payload (e.g., an IP packet), not another MPLS label. This is a fundamental part of the MPLS data plane that SR reuses.

VLAN-aware bundling allows multiple VLANs (a bundle) to be mapped to a single EVPN instance (EVI) and a single service label (per PE). This is in contrast to VLAN-based service, which requires a separate EVI and label per VLAN. For hundreds of VLANs, VLAN-aware bundling dramatically scales down the control plane (fewer EVPN MAC/IP Advertisement and Inclusive Multicast Ethernet Tag routes) and the data plane (fewer service labels), which is a key design consideration for large-scale deployments like DCI.

IS-IS is a link-state protocol. The metric carried in the LSP for a prefix (like a loopback) is the metric from the originating router to that prefix. By default, this is the internal IGP metric of the link to the loopback, often 0. When this LSP is flooded and processed by another router, that router calculates the total cost to reach the prefix by adding the cost from itself to the advertising router (the link cost, here 10) to the metric found in the LSP (100). The result is the total path cost (110) installed in the RIB.

Equal-Cost Multi-Path (ECMP) routing is the feature that allows a router to install multiple next-hops for a single prefix in the FIB and to distribute traffic across them. If ECMP is not enabled or is misconfigured (e.g., a low `ecmp` value in the config router context), the router will select only one best next-hop, even if multiple equal-cost paths exist in the RIB. Enabling and properly configuring ECMP is essential for utilizing all available bandwidth and for redundancy.

The fundamental purpose of a standby secondary LSP is to provide a pre-signaled backup path. The primary trigger for an automatic switchover is the failure of the primary LSP, meaning its operational state transitions to `down`. This is typically detected via loss of RSVP hello messages or path/resv errors. The switchover is then handled by the fast-reroute or standby signaling mechanisms. Other options like metric changes or bandwidth reductions do not necessarily bring the LSP down and would not trigger an automatic switch to a standby.

BGP’s default hold timer is 90 seconds. A physical or data-link failure could take up to 90 seconds to be detected by BGP, causing a long blackhole period. BFD (Bidirectional Forwarding Detection) is a lightweight, rapid failure detection protocol. By configuring aggressive BFD timers (e.g., 50ms transmit/50ms receive with a multiplier of 3), BFD can detect a failure in ~150ms. BGP can then be notified to tear down its session almost immediately, drastically reducing convergence time for critical peers, which is essential for inter-AS links.

In the Rosen MVPN model (now often called “Default MDT”), a Multicast Domain (MD) is created in the service provider core. This MD is a separate routing instance (e.g., a VRF) that runs its own PIM protocol. Customer VPN multicast traffic is encapsulated into GRE/mGRE tunnels and transported within this MD. The MD’s PIM control plane (PIM neighborships, joins, prunes) is what establishes the provider tunnels (P-tunnels) between PEs, completely isolating the customer multicast control plane from the core’s unicast IGP/BGP.

For a Flex-Algo to work, all routers that could potentially be part of a path computed under that algorithm must understand and advertise the relevant link attributes. If a router does not participate (e.g., does not advertise its TE Metric or link color exclusions), it creates a ‘black hole’ in the topology database for that Flex-Algo. The head-end router’s CSPF calculation would be incomplete or incorrect. Therefore, participation is required network-wide, or at minimum, within the domain where the Flex-Algo is intended to be used.

BGP runs over TCP (port 179). Stateful firewalls often have configurable idle timeouts for TCP sessions. If the firewall’s TCP timeout is shorter than BGP’s keepalive/hold timers, the firewall will tear down its state table entry, causing subsequent BGP packets (like keepalives) to be dropped, breaking the session. The peers will then restart the TCP connection, repeating the cycle. This is a classic real-world scenario distinct from simple configuration errors on the routers themselves.0