xDev Asia

Jobs (batch tasks), CronJobs (scheduled tasks). Resource requests và limits, LimitRange, ResourceQuota. QoS classes cho CKAD exam.

Jobs và CronJobs — completions, parallelism, concurrencyPolicy

1. Jobs

Một Job tạo một hoặc nhiều Pods và đảm bảo chúng hoàn thành thành công. Khi Job hoàn thành, Pods không bị xóa (cho log inspection).

apiVersion: batch/v1
kind: Job
metadata:
  name: data-processor
spec:
  completions: 3       # Run 3 successful completions
  parallelism: 2       # Run 2 pods at a time
  backoffLimit: 4      # Retry up to 4 times on failure
  template:
    spec:
      restartPolicy: Never  # OnFailure or Never (required for Job)
      containers:
      - name: processor
        image: busybox
        command: ['sh', '-c', 'echo Processing; sleep 5']

Field	Ý nghĩa	Default
`completions`	Số lần phải hoàn thành	1
`parallelism`	Số Pods chạy concurrent	1
`backoffLimit`	Số lần retry khi fail	6
`activeDeadlineSeconds`	Timeout tổng thể của Job	Unlimited

Exam tip: Job Pods phải có restartPolicy: Never hoặc OnFailure. Không thể dùng Always (default cho regular Pods). CKAD hay test việc tạo Job và kiểm tra completion status.

2. CronJobs

apiVersion: batch/v1
kind: CronJob
metadata:
  name: nightly-backup
spec:
  schedule: "0 2 * * *"  # Cron syntax: minute hour day month weekday
  concurrencyPolicy: Forbid  # Allow | Forbid | Replace
  successfulJobsHistoryLimit: 3
  failedJobsHistoryLimit: 1
  jobTemplate:
    spec:
      template:
        spec:
          restartPolicy: OnFailure
          containers:
          - name: backup
            image: backup-tool:1.0
            command: ['./backup.sh']

concurrencyPolicy	Hành vi
Allow	Cho phép Jobs chạy concurrent (default)
Forbid	Skip new Job nếu previous chưa xong
Replace	Cancel previous Job, start new one

3. Resource Requests & Limits

spec:
  containers:
  - name: app
    image: myapp
    resources:
      requests:
        cpu: "250m"      # 0.25 CPU core (minimum guaranteed)
        memory: "128Mi"  # 128 MiB minimum
      limits:
        cpu: "500m"      # Max 0.5 CPU core
        memory: "256Mi"  # Max 256 MiB (OOM if exceeded)

QoS Classes (dựa trên requests/limits):

Guaranteed:  requests == limits (both CPU and memory)
             → Last to be evicted under pressure

Burstable:   requests < limits (or only one set)
             → Middle priority for eviction

BestEffort:  NO requests, NO limits
             → First to be evicted under pressure

Exam tip: Để Pod có QoS class Guaranteed: phải set cả cpu và memory trong cả requests và limits, và chúng phải bằng nhau. Mỗi container trong Pod đều phải thỏa mãn điều kiện này.

4. LimitRange & ResourceQuota

Object	Scope	Mục đích
LimitRange	Namespace	Set default requests/limits cho Pods/Containers trong namespace
ResourceQuota	Namespace	Giới hạn tổng resources namespace được dùng

ResourceQuota example:
apiVersion: v1
kind: ResourceQuota
metadata:
  name: dev-quota
  namespace: development
spec:
  hard:
    requests.cpu: "4"
    requests.memory: "8Gi"
    limits.cpu: "8"
    limits.memory: "16Gi"
    pods: "20"

5. Cheat Sheet

Câu hỏi exam	Đáp án
Job cần restartPolicy gì?	`Never` hoặc `OnFailure`
CronJob mỗi 5 phút?	`/5 * * *`
Container bị OOM Kill do gì?	Vượt `limits.memory`
QoS Guaranteed cần gì?	requests == limits (cả CPU và Memory)
Giới hạn resources của namespace?	ResourceQuota

6. Practice Questions

Q1: A Job is configured with completions: 5 and parallelism: 2. How does it execute?

A) Creates 5 Pods simultaneously until all complete
B) Runs 2 Pods at a time, creating new ones as old ones complete, until 5 total completions ✓
C) Runs 5 Pods sequentially one by one
D) Creates 2 Pods, each completing 2.5 times

Explanation: completions=5 means 5 PODs must exit successfully. parallelism=2 means at most 2 run at once. As each Pod completes, a new one starts until completion count is reached. Total Pods created could be more if some fail.

Q2: A Pod has no resource requests or limits set. What QoS class is it assigned and how does this affect eviction?

A) Guaranteed — it will be last to be evicted
B) Burstable — it has medium eviction priority
C) BestEffort — it will be first to be evicted under resource pressure ✓
D) NoQoS — it has no eviction priority

Explanation: Pods with no resource requests or limits get BestEffort QoS class. When nodes face resource pressure, Kubernetes evicts BestEffort Pods first to free resources for higher-priority workloads.

Q3: A CronJob is scheduled every hour. A previous job is still running when the next scheduled time arrives. With concurrencyPolicy: Forbid, what happens?

A) The running job is cancelled; the new one starts
B) Both jobs run concurrently
C) The new job is skipped; the running job continues ✓
D) The CronJob is suspended until the running job completes