High Availability, Operations, and Troubleshooting
This domain tests keeping the fabric running and diagnosing failures. Know the HA building blocks: redundant power and fans, VSX active-active pairs, dual-partition boot for safe upgrades, and checkpoint/rollback for configuration recovery. Understand that distributed services add a dimension - DPU state and policy distribution must be considered alongside traditional forwarding HA.
Operational fluency matters: upgrade workflow (stage, reboot to alternate partition, verify, commit), configuration backup, and health monitoring via telemetry and logs. Troubleshooting should follow a layered method: physical/link, L2, L3 underlay, overlay (EVPN/VXLAN), then services (DPU policy/state).
Traps:
- Assuming VSX failover preserves stateful DPU sessions across a switch failure - plan for session re-establishment.
- Jumping to policy changes before confirming underlay/overlay health.
- Forgetting that a failed upgrade can be recovered by booting the prior partition.
- Ignoring asymmetric routing as a cause of "random" service drops.
Rule of thumb: troubleshoot bottom-up - link, then L2, then underlay, then overlay, then services. Worked example: intermittent drops between tiers. Check interface errors (link), VLAN/VNI mapping (L2/overlay), VTEP loopback reachability (underlay), EVPN Type 2 presence (overlay), then DPU session tables and policy counters (services). If both directions don't hit the same DPU, fix routing symmetry. Verify each layer with its native show command before moving up.
Also review how to collect logs and telemetry for escalation.