Troubleshooting and High Availability
The final domain combines structured fault isolation with redundancy design: proving whether a fault sits at Layer 1, 2 or 3, and confirming that VLT, VRRP, LACP and routing redundancy behave as intended.
A workable method
Start with physical and interface state, then Layer 2, then Layer 3, then application. Useful commands on OS10 include show interface status, show interface counters, show mac address-table, show lldp neighbors, show spanning-tree, show vlt mismatch, show ip route, show ip ospf neighbor and show logging. Change one variable at a time and clear counters only after recording the baseline.
Reading counters
CRC and alignment errors indicate cabling, connectors or optics. Runts and giants point at MTU or frame corruption. Late collisions indicate duplex mismatch. Output discards suggest congestion or QoS policing rather than a bad cable. An interface that is up/up can still be losing traffic.
High availability
VLT gives dual-homed servers an active/active path; verify domain ID, VLTi and backup link health. VRRP provides a gateway virtual IP; confirm master/backup roles and that both can reach upstream. LACP checks member state and hashing; BFD and graceful restart speed convergence. Use ISSU or staged reloads in maintenance windows so redundant paths carry traffic during upgrades.
Worked example
Users on one rack report intermittent loss to a server. Interface counters on the server-facing port show CRC errors climbing while discards stay flat, and LLDP confirms the expected peer. That pattern is Layer 1: replace the DAC/optics and re-test. If CRC were clean but output discards rose during backups, the fault would be congestion/oversubscription, not the cable - a different fix entirely.
Traps
Accepting up/up as healthy; clearing counters before capturing evidence; replacing hardware before reading the error type; assuming VLT protects against every failure including a shared upstream; forgetting that an unsaved fix disappears at the next reload.