Cheat sheet & self-check

Pick an answer to see if you got it, and why.

1. Q1. What is a BMC?

   1. AA Kubernetes controller
   2. BA small independent computer on the server board (iDRAC, iLO, XClarity, etc.) that controls power, boot and console even when the main OS is off
   3. CA network switch
   4. DA type of disk
   Show answer

   B. BMCs make remote provisioning possible. Their network must be isolated and their default credentials changed.

   From lesson 04 · Bare-metal provisioning

2. Q2. Why is Redfish preferred over IPMI for automation?

   1. AIPMI is faster
   2. BRedfish is a modern, standard HTTPS/JSON REST API with richer features (virtual media, firmware updates, inventory), while IPMI is older, less secure and inconsistent
   3. CRedfish needs no network
   4. DIPMI can't power servers
   Show answer

   B. Most provisioning tools (Tinkerbell's Rufio, Metal3/Ironic) support Redfish first; IPMI remains for old hardware.

   From lesson 04 · Bare-metal provisioning

3. Q3. UEFI servers PXE-boot, but legacy-BIOS servers on the same network don't. Likely cause?

   1. ABad cables
   2. BThe DHCP server hands out a UEFI binary to everyone; it must choose the boot file by client architecture (option 93)
   3. CWrong Kubernetes version
   4. DToo many VLANs
   Show answer

   B. BIOS needs a legacy PXE binary (e.g. undionly.kpxe); UEFI needs an EFI binary (e.g. ipxe.efi).

   From lesson 04 · Bare-metal provisioning

4. Q4. In Tinkerbell, what does a Workflow combine?

   1. ATwo clusters
   2. BA Template (the steps) with a specific Hardware record (the machine), so the machine's worker runs those steps
   3. CA Deployment and a Service
   4. DOnly DHCP settings
   Show answer

   B. Templates are reusable; workflows bind them to machines. Each action is a container image run by the worker on the in-memory OS.

   From lesson 05 · Tinkerbell & Metal3

5. Q5. What does Metal3's BareMetalHost resource represent?

   1. AA virtual machine
   2. BA physical server with its BMC address, credentials, boot MAC, and the image to provision, managed by the Bare Metal Operator through Ironic
   3. CA pod
   4. DA network switch
   Show answer

   B. The operator inspects, cleans and provisions the host to match the spec.

   From lesson 05 · Tinkerbell & Metal3

6. Q6. How do Tinkerbell and Metal3 relate to Cluster API?

   1. AThey replace it
   2. BEach has a Cluster API infrastructure provider (CAPT, CAPM3), so CAPI Machines are backed by provisioned bare-metal servers
   3. CThey're unrelated
   4. DOnly Metal3 works with CAPI
   Show answer

   B. CAPI handles the cluster lifecycle; the provider turns 'I need a machine' into 'provision that server'.

   From lesson 05 · Tinkerbell & Metal3

7. Q7. Why use immutable, versioned OS images for edge nodes?

   1. AThey're smaller
   2. BEvery node boots the exact same tested bits; changes ship as a new image version (tested once, rolled out everywhere) instead of drifting per-node patches
   3. CThey need no updates
   4. DThey can't be hacked
   Show answer

   B. Reproducibility is what makes hundreds of sites manageable and rollbacks possible.

   From lesson 06 · OS image creation

8. Q8. What's the role of cloud-init in a ZTP flow?

   1. ABuilding the image
   2. BApplying per-node settings at first boot (hostname, network, users/SSH keys, files, join commands) from user-data and metadata, so one image serves every node
   3. CReplacing Kubernetes
   4. DManaging BMCs
   Show answer

   B. The image is generic; cloud-init (or Ignition) personalises it at boot.

   From lesson 06 · OS image creation

9. Q9. Why is RAW often used for bare-metal provisioning?

   1. AIt's encrypted
   2. BIt can be streamed directly onto the disk block by block (compressed in transit), with no installer step
   3. CIt's the only format BIOS understands
   4. DIt's smaller than QCOW2 uncompressed
   Show answer

   B. Image-to-disk streaming is fast and deterministic. Compress it (gz/xz) for transfer.

   From lesson 06 · OS image creation

10. Q10. Why mirror images by digest and with --all architectures?

    1. AIt's faster
    2. BDigests guarantee the exact tested bytes arrive at the site, and --all keeps multi-arch manifests intact so nodes of any architecture can pull
    3. CRegistries require it
    4. DTo save disk
    Show answer

    B. A tag can move between download and use; a partial manifest breaks nodes of another architecture.

    From lesson 07 · Air-gapped delivery

11. Q11. What belongs in a release bundle for air-gapped sites?

    1. AOnly container images
    2. BEverything for one tested release: images, Helm charts, OS images, manifests and cluster specs, with a manifest of versions and digests, checksums and signatures
    3. COnly the Git repo
    4. DJust the Kubernetes binary
    Show answer

    B. A site should be able to install or upgrade from the bundle alone.

    From lesson 07 · Air-gapped delivery

12. Q12. How do you stop nodes silently pulling from the internet?

    1. AHope
    2. BConfigure every image source to use the internal mirror (containerd/distribution mirror settings, system-default-registry), and block egress, so a missing image fails loudly during testing
    3. CDelete Docker Hub
    4. DUse latest tags
    Show answer

    B. Test the bundle in a truly disconnected environment before shipping it.

    From lesson 07 · Air-gapped delivery