Switching, routing, and multicast diagnosis
Begin with the intended forwarding path
Draw the source, destination, VLAN, routing context, gateway, and physical dependencies before changing a device. Management access, a routing session, and an application session are different observations. A successful small ping does not validate a large encapsulated packet, an allowed application port, or the return path. State what each test proves and what it leaves open.
For a trunk or port-channel, compare intended and operational member state, allowed VLANs, native behavior, and relevant consistency parameters. Aggregate bandwidth is not a promise that a single hashed flow uses every member. Check the platform's hashing behavior and traffic diversity before diagnosing imbalance from one conversation. Distinguish physical redundancy from independent failure domains: links can share a line card, power source, tray, or upstream device.
Diagnose vPC and spanning tree with their actual roles
The vPC peer-link and peer-keepalive have different functions. In the normal peer-link-loss scenario where the other peer remains alive, secondary-side protection prevents unsafe dual-active forwarding. Use the relevant show commands to examine peer state, member state, consistency, and the associated VLANs. Do not treat keepalive as a replacement user-data path or assume that a healthy management route proves the configured keepalive VRF works.
Spanning-tree interpretation depends on topology and port roles. Edge treatment belongs on an appropriate endpoint attachment, not automatically every link that should converge quickly. BPDU guard and other protections enforce specific assumptions. MST region identity requires compatible configuration, not merely matching one instance number.
Separate routing adjacency from route usability
OSPF neighbors require compatible parameters and network-type expectations. A two-way state between broadcast-network DROTHERs can be normal, while EXSTART after a link change warrants examination of database exchange and MTU. An ignore option does not certify the underlying packet-size path. Area types must match redistribution requirements.
BGP session establishment does not guarantee that all required prefixes are accepted, reflected, installed, or have reachable next hops. Compare selection attributes in the correct order and distinguish local Cisco weight from propagated local preference. Prefix policies must specify intended length matching. Mutual redistribution needs explicit boundary and loop-prevention policy. Longest-prefix lookup selects among usable installed routes and should not be confused with choosing protocol preference for the same prefix.
For multicast, connect receiver membership, querier behavior, rendezvous requirements, reverse-path checks, and distribution state. A valid unicast route or one receiver report is not the entire multicast forwarding design. Validate the specific tree and traffic direction with representative evidence.