Init Container Management

Resource sharing within containers

Given the order of execution for init, sidecar and app containers, the following rules for resource usage apply:

  • The highest of any particular resource request or limit defined on all init containers is the effective init request/limit. If any resource has no resource limit specified this is considered as the highest limit.
  • The Pod's effective request/limit for a resource is the sum of pod overhead and the higher of:
  • the sum of all non-init containers(app and sidecar containers) request/limit for a resource
  • the effective init request/limit for a resource
  • Scheduling is done based on effective requests/limits, which means init containers can reserve resources for initialization that are not used during the life of the Pod.
  • Effective init = highest among init containers
  • Pod’s effective request = max(sum(non-init containers), effective init) + overhead
  • Scheduling, quota, QoS decisions use the effective pod request/limit
  • Init can reserve resources even if not used later

🔥 Exam-Style Questions

Q1

A Pod has:

  • Two init containers:

  • init-A: 200 Mi memory request

  • init-B: 150 Mi memory request

  • Two app containers (non-init):

  • app-1: 120 Mi request

  • app-2: 140 Mi request

Assume no overhead. What is the Pod’s effective memory request (for scheduling)? If the namespace has a ResourceQuota of 300 Mi, will this Pod be allowed?

Answer

  • Effective init = max(200, 150) = 200 Mi
  • Sum non-init = 120 + 140 = 260 Mi
  • Effective Pod request = max(200, 260) = 260 Mi
  • Quota is 300 Mi, so yes, this Pod fits (260 ≤ 300).

Q2

You have a Deployment (1 replica) with:

  • Two init containers:

  • init-setup1: 300 Mi memory request

  • init-setup2: 500 Mi memory request

  • Two application containers:

  • frontend: 250 Mi

  • backend: 200 Mi

Also assume a pod overhead of 50 Mi. Compute the effective memory request. If the namespace quota is 900 Mi, is this Pod allowed?

Answer

  • Effective init = max(300, 500) = 500 Mi
  • Sum non-init = 250 + 200 = 450 Mi
  • Pod overhead = 50 Mi → we add overhead (if asked)
  • Effective Pod request = max(500, 450) + 50 = 550 Mi
  • Namespace quota = 900 Mi → yes, 550 ≤ 900, so allowed.

Q3

A Deployment with 3 replicas. For each Pod:

  • init containers:

  • init1: 400 Mi

  • init2: 600 Mi

  • application containers:

  • serviceA: 350 Mi

  • serviceB: 250 Mi

No overhead.

  1. What is effective memory per Pod?
  2. Total across all replicas?
  3. If namespace quota is 1800 Mi, will the Deployment be admitted?

Answer

  • Effective init = max(400, 600) = 600 Mi
  • Sum non-init = 350 + 250 = 600 Mi
  • Effective Pod request = max(600, 600) = 600 Mi
  • 3 replicas × 600 = 1800 Mi total
  • Namespace quota = 1800 Mi → it exactly matches, so okay (barely).

Q4

Pod has:

  • init containers:

  • init1: request = 100 Mi

  • init2: no memory limit/request specified

  • app containers:

  • worker: 120 Mi

  • logger: 30 Mi

What is Pod’s effective memory request (consider “no limit / no request” in init as highest)?

Answer

  • init1 has 100 Mi request
  • init2 has no memory request or limit specified → this is considered highest (unbounded) or effectively infinity for that resource
  • So effective init = “no limit” / effectively infinite → meaning init could demand more
  • Sum non-init = 120 + 30 = 150 Mi
  • Pod’s effective = whichever is higher: effective init (unbounded) vs 150 → effective init wins
  • So Pod effective request = “unbounded / infinite” — in practice that means scheduling fails or is constrained by node or quota. (In real K8s, lacking a limit is “best-effort”, but for init container lacking explicit request is considered highest)

Q5

A Deployment with 2 replicas. Each Pod:

  • init containers:

  • initA: 150 Mi

  • initB: 250 Mi

  • app containers:

  • web: 200 Mi

  • api: 180 Mi

Also consider pod overhead of 20 Mi. Compute effective request per Pod and total for the Deployment. If namespace quota = 900 Mi, is it okay?

Answer

  • Effective init = max(150, 250) = 250 Mi
  • Sum non-init = 200 + 180 = 380 Mi
  • Pod overhead = 20 Mi → adding if needed
  • Effective Pod request = max(250, 380) + 20 = 400 Mi
  • Deployment has 2 replicas → total = 2 × 400 = 800 Mi
  • Namespace quota = 900 Mi → 800 ≤ 900, so acceptable.