Cloud-Native Architecture: Containers, Schedulers and the Control Loop · seed 1 · A4, ink-friendly. The answer key prints on its own page for grown-ups.

Declare the state, let the loop converge

Computing · Computer Systems · ages 23-24
Name ______________________   Date ____________

  1. In the Linux container model taught here, which description of isolation and sharing is accurate?

    • Selected process views are isolated; containers share the host kernel.
    • Selected process views are shared; each container receives a separate kernel.
    • Selected process views are shared; containers share the host kernel.

  2. Orders and payments are separately deployed services that communicate over a network. Which architecture description fits?

    • One monolithic program
    • One local-call program
    • Separate microservices

  3. A monolithic application can run in a container in the cloud.

    Circle one:   True   False

  4. A workload needs 2 GiB of available memory and must meet a placement rule. A meets the rule with 1 GiB; B meets it with 4 GiB; C breaks it with 8 GiB. Which node meets both stated conditions?

    • A's placement match makes its 1 GiB sufficient.
    • B satisfies both conditions.
    • C's 8 GiB makes its placement violation irrelevant.

  5. After failure cleanup, a controller observes 2 managed replica objects against a target of 5. No creation is pending and no concurrent changes occur. How many more objects should it request?

    • 3 additional objects
    • 5 additional objects
    • 2 additional objects

  6. A failure changes the current replica count. What does a standing desired-state target provide that a one-time command list alone does not?

    • Guaranteed startup after each failure, regardless of capacity.
    • Permission to skip later checks of existing replica objects.
    • A target for comparing state and repairing later drift.

  7. When its managed replica count matches the target, a controller should create an extra replica merely because another reconciliation pass has begun.

    Circle one:   True   False

  8. A controller creates replica object worker-b but loses the success response. On retry, it confirms worker-b exists and the target is met. What should it do?

    • Create worker-c on retry.
    • Keep worker-b and avoid duplicating the completed creation.
    • Delete worker-b so that the retry always begins with no replica object.
LightMySky · lightmysky.comW1-mt_nV1LgSRsh_-s1

Answer key

For grown-ups. Fold this page away before handing over the rest.

Declare the state, let the loop converge W1-mt_nV1LgSRsh_-s1

  1. Selected process views are isolated; containers share the host kernel. · Namespaces isolate selected views, while containers use the host kernel. An image does not include a separate kernel for each container.
  2. Separate microservices · The stated deployment units are separate networked services. That fits microservices, without proving that their failures cannot affect one another.
  3. True · A monolith is deployed as one application unit. That unit can still be packaged in a container and hosted in the cloud.
  4. B satisfies both conditions. · B meets the rule and has at least 2 GiB available. A lacks enough memory; C violates a required rule despite having more memory.
  5. 3 additional objects · Five are wanted and two exist, so the shortfall is 5 - 2 = 3. Creating objects does not establish that their processes are ready.
  6. A target for comparing state and repairing later drift. · The target remains available for later comparisons and repair decisions. It does not remove capacity, configuration or startup limits.
  7. False · A new pass calls for checking state, not automatically adding replicas. Matching this count needs no additional replica creation.
  8. Keep worker-b and avoid duplicating the completed creation. · The missing response did not undo creation. The retry has confirmed that the object exists and the target is met, so no duplicate replica is needed.
Worksheet · LightMySky