Installation and Configuration
This domain covers how a host travels from bare metal to a validated, production-ready node: hardware compatibility, firmware state, boot-mode choice, network identity, time synchronization, and the post-install checks that prove the build matches the design. Questions here test planning and verification, not memorised installer screens.
Before the installer runs
- Confirm every component - CPU generation, NIC, HBA and disk model - appears on the vendor compatibility matrix, not just the server family name.
- Bring BIOS and device firmware to a supported baseline; mismatched firmware is the single most common cause of failed or unstable builds.
- Decide UEFI versus legacy BIOS and keep that choice identical across the cluster; mixed modes complicate boot ordering and PXE.
- Verify the image checksum (SHA-256) before writing media. A corrupt ISO typically fails deep into the install, wasting hours.
- Record the design: management IP, gateway, DNS servers, NTP sources, FQDN and VLAN IDs.
During and after
Assign a static management address and an FQDN with matching forward and reverse records. Point NTP at at least two sources and confirm the offset stays under a second, because certificate validation and directory authentication reject skewed clocks. Then validate: all links up at the expected speed and duplex, DNS resolving in both directions, time synchronized, and no outstanding alerts.
Worked example
A node installs cleanly but cluster creation fails on time synchronization. Reasoning: the hostname resolves and the NTP service is running, so the fault is reachability - an unopened UDP 123 rule on the management VLAN. Confirm with chronyc sources (no reachable peers), open the rule, then re-verify the offset before retrying the join.
Traps
Skipping checksum verification; leaving one node in legacy boot while the rest are UEFI; omitting PTR records; assuming DHCP is acceptable for management interfaces; never checking duplex mismatches on older switches.