Data Center Operations, Support, and Troubleshooting
This domain tests day-two work: firmware and configuration management, health monitoring, log collection, and a structured troubleshooting path from physical layer to application. Questions often describe a symptom and ask for the first check or the correct escalation artifact.
Operations
- UCS firmware bundles combine FI, IOM, blade BIOS, CIMC, and adapter firmware; check release notes for the supported upgrade path and for Auto Install versus manual upgrade.
- Back up UCS Manager configuration (full state and all configuration) before upgrades; keep a change log and label both cable ends.
- Health checks: PSU, fan, temperature, and SEL; CIMC and UCS Manager fault tabs; Nexus show environment, show inventory, and show interface.
- Support: capture show tech-support, core files, and serial console output; open a TAC case with serial number and firmware version; follow the RMA process.
Troubleshooting layers
- Physical: link LED, cable, optic, PSU. Link down with no light often means wrong optic, damaged cable, or a disabled port.
- Discovery: a blade missing from UCS Manager often means the chassis discovery policy (port-channel versus discrete) does not match the IOM cabling, or FI ports are unconfigured.
- Boot: check boot policy order, SAN zoning and LUN masking, and POST messages; a server seeing no boot device may have lost its vHBA login.
- Logs: correlate timestamps and fix the first fault, not the cascade.
Worked example
A new B200 M6 blade appears in the chassis but not in UCS Manager. The IOMs are cabled to FI ports 1-4 and 5-8, but the chassis discovery policy is set to discrete while the cabling is port-channel. Fix: set discovery to port-channel (or recable to match), acknowledge the chassis, then verify FI port state and licensing. Reasoning: the discovery policy and IOM-to-FI cabling must agree before the chassis is managed.
Traps
- Upgrading without a backup or a checked upgrade path.
- Treating the first console error as root cause when an earlier PSU or SEL event started the cascade.
- Forgetting that some service-profile changes need a reboot to apply.
- Assuming link up means the optic is correct; it may be the wrong speed or type.