# DevOps 인터뷰 필수 질문: 완전 가이드 2026

> CI/CD, Kubernetes, Docker, Terraform, SRE 실천에 관한 필수 질문으로 DevOps 인터뷰를 준비하십시오. 상세한 답변 포함.

- Published: 2026-01-23
- Updated: 2026-04-10
- Author: SharpSkill
- Tags: devops, interview, ci cd, kubernetes, infrastructure
- Reading time: 15 min

---

DevOps 인터뷰는 개발, 운영, 자동화 문화 기술의 독특한 조합을 평가합니다. 이 가이드는 도메인별로 정리된 가장 자주 묻는 질문들을 개념에 대한 깊은 숙련도를 보여주는 구조화된 답변과 함께 엮었습니다.

> **준비 팁**
>
> 기술적 지식 외에도, 면접관은 복잡한 개념을 간단하게 설명하는 능력과 구체적인 문제 해결 경험을 공유할 수 있는지를 평가합니다.

## DevOps 기초와 문화

초기 질문은 DevOps 철학에 대한 전반적인 이해를 평가하는 경우가 많습니다.

**Q1: DevOps란 무엇이며 이 접근 방식은 어떤 문제를 해결합니까?**

DevOps는 소프트웨어 개발(Dev)과 IT 운영(Ops)을 통합하는 문화와 실천의 집합입니다. 이 접근 방식은 높은 품질을 유지하면서 개발 주기를 단축하는 것을 목표로 합니다.

```yaml
# devops-principles.yaml
# The pillars of DevOps culture
principles:
  collaboration:
    description: "Breaking silos between teams"
    practices:
      - "Shared responsibility for production code"
      - "Continuous communication via ChatOps"
      - "Blameless post-mortems"

  automation:
    description: "Automate repetitive tasks"
    practices:
      - "Infrastructure as Code (IaC)"
      - "CI/CD pipelines"
      - "Automated testing at all levels"

  measurement:
    description: "Measure to improve"
    metrics:
      - "Deployment frequency"
      - "Lead time for changes"
      - "Mean time to recovery (MTTR)"
      - "Change failure rate"

  sharing:
    description: "Share knowledge"
    practices:
      - "Documentation as Code"
      - "Automated runbooks"
      - "Regular knowledge sharing sessions"
```

해결되는 문제에는 느리고 위험한 배포, 팀 간 가시성 부족, 환경 간 불일치 등이 포함됩니다.

**Q2: CI, CD(Continuous Delivery), CD(Continuous Deployment)의 차이점은 무엇입니까?**

이 세 가지 개념은 배포 주기 자동화의 점진적 발전을 형성합니다.

```bash
# ci-cd-pipeline-stages.sh
# Illustration of CI/CD stages

# ============================================
# CI (Continuous Integration)
# ============================================
# Goal: Frequently integrate code into a shared repository
# Automation: Build + Tests
echo "CI: Code commit → Build → Unit Tests → Integration Tests"

# ============================================
# CD (Continuous Delivery)
# ============================================
# Goal: Code always deployable to production
# Automation: CI + Staging deployment + Manual approval
echo "CD Delivery: CI → Deploy Staging → Manual Approval → Deploy Prod"

# ============================================
# CD (Continuous Deployment)
# ============================================
# Goal: Automatic deployment to production
# Automation: Entire pipeline without human intervention
echo "CD Deployment: CI → Deploy Staging → Auto Tests → Auto Deploy Prod"
```

핵심 차이점은 자동화 수준에 있습니다. Continuous Delivery는 프로덕션 배포 전 수동 승인이 필요하지만, Continuous Deployment는 전체 프로세스를 완전히 자동화합니다.

## CI/CD와 파이프라인

CI/CD 질문은 배포 파이프라인 설계 및 최적화 능력을 테스트합니다.

**Q3: 견고한 CI/CD 파이프라인을 어떻게 구성합니까?**

잘 설계된 파이프라인은 각 레벨에 체크포인트가 있는 단계적 스테이지를 따릅니다.

```yaml
# .gitlab-ci.yml
# Complete CI/CD pipeline with parallel and sequential stages
stages:
  - validate
  - build
  - test
  - security
  - deploy-staging
  - integration-tests
  - deploy-production

variables:
  DOCKER_IMAGE: $CI_REGISTRY_IMAGE:$CI_COMMIT_SHA

# ============================================
# Stage 1: Fast validation (< 2 min)
# ============================================
lint:
  stage: validate
  script:
    - npm run lint
    - npm run type-check
  # Run on every commit
  rules:
    - if: $CI_PIPELINE_SOURCE == "merge_request_event"
    - if: $CI_COMMIT_BRANCH

# ============================================
# Stage 2: Application build
# ============================================
build:
  stage: build
  script:
    - docker build -t $DOCKER_IMAGE .
    - docker push $DOCKER_IMAGE
  # Cache Docker layers to speed up builds
  cache:
    key: docker-$CI_COMMIT_REF_SLUG
    paths:
      - .docker-cache/

# ============================================
# Stage 3: Parallel tests
# ============================================
unit-tests:
  stage: test
  script:
    - npm run test:unit -- --coverage
  coverage: '/Lines\s*:\s*(\d+\.?\d*)%/'
  artifacts:
    reports:
      coverage_report:
        coverage_format: cobertura
        path: coverage/cobertura-coverage.xml

integration-tests:
  stage: test
  services:
    - postgres:16-alpine
    - redis:7-alpine
  script:
    - npm run test:integration
  # Parallelization with unit tests
  parallel: 3

# ============================================
# Stage 4: Security analysis
# ============================================
sast:
  stage: security
  script:
    - trivy image --exit-code 1 --severity HIGH,CRITICAL $DOCKER_IMAGE
  allow_failure: false

dependency-scan:
  stage: security
  script:
    - npm audit --audit-level=high
  allow_failure: true  # Alert without blocking

# ============================================
# Stage 5: Staging deployment
# ============================================
deploy-staging:
  stage: deploy-staging
  script:
    - kubectl set image deployment/app app=$DOCKER_IMAGE -n staging
    - kubectl rollout status deployment/app -n staging --timeout=300s
  environment:
    name: staging
    url: https://staging.example.com
  only:
    - develop

# ============================================
# Stage 6: E2E tests on staging
# ============================================
e2e-tests:
  stage: integration-tests
  script:
    - npm run test:e2e -- --base-url=https://staging.example.com
  artifacts:
    when: on_failure
    paths:
      - cypress/screenshots/
      - cypress/videos/
  only:
    - develop

# ============================================
# Stage 7: Production deployment
# ============================================
deploy-production:
  stage: deploy-production
  script:
    - kubectl set image deployment/app app=$DOCKER_IMAGE -n production
    - kubectl rollout status deployment/app -n production --timeout=300s
  environment:
    name: production
    url: https://app.example.com
  # Manual deployment with protection
  when: manual
  only:
    - main
```

이 파이프라인은 모범 사례를 보여줍니다: 속도를 위한 병렬 스테이지, 추적성을 위한 아티팩트, 프로덕션을 위한 보호된 환경.

**Q4: CI/CD 파이프라인에서 시크릿을 어떻게 관리합니까?**

시크릿 관리는 암호화, 교체, 최소 권한 원칙을 결합한 다층 접근이 필요합니다.

```yaml
# kubernetes-secrets-management.yaml
# Approach 1: External Secrets Operator with HashiCorp Vault
apiVersion: external-secrets.io/v1beta1
kind: ExternalSecret
metadata:
  name: app-secrets
  namespace: production
spec:
  refreshInterval: 1h  # Automatic rotation
  secretStoreRef:
    name: vault-backend
    kind: ClusterSecretStore
  target:
    name: app-secrets
    creationPolicy: Owner
  data:
    # Reference to secret in Vault
    - secretKey: DATABASE_PASSWORD
      remoteRef:
        key: secret/data/production/database
        property: password
    - secretKey: API_KEY
      remoteRef:
        key: secret/data/production/api
        property: key
---
# SecretStore configuration
apiVersion: external-secrets.io/v1beta1
kind: ClusterSecretStore
metadata:
  name: vault-backend
spec:
  provider:
    vault:
      server: "https://vault.example.com"
      path: "secret"
      version: "v2"
      auth:
        kubernetes:
          mountPath: "kubernetes"
          role: "external-secrets"
          # Dedicated ServiceAccount with minimal permissions
          serviceAccountRef:
            name: external-secrets-sa
```

권장 사례: 코드에 시크릿을 평문으로 저장하지 않기, 전용 시크릿 매니저(Vault, AWS Secrets Manager) 사용, 자동 교체 활성화.

> **안티패턴**
>
> 로그에 보이는 CI/CD 환경 변수를 피하십시오. CI 플랫폼의 기본 기능(마스크된 변수)을 사용해 항상 시크릿을 마스킹하십시오.

## Kubernetes와 오케스트레이션

Kubernetes 질문은 오케스트레이션 개념에 대한 이해와 구체적인 문제 해결 능력을 평가합니다.

**Q5: Kubernetes 아키텍처와 각 컴포넌트의 역할을 설명하십시오.**

Kubernetes는 서로 다른 책임을 가진 컴포넌트로 구성된 마스터-노드 아키텍처를 따릅니다.

```yaml
# kubernetes-architecture.yaml
# Control Plane components (Master)
control_plane:
  api_server:
    role: "Entry point for all API requests"
    responsibilities:
      - "Validation and configuration of API objects"
      - "Authentication and authorization"
      - "REST interface for kubectl and other clients"

  etcd:
    role: "Distributed key-value database"
    responsibilities:
      - "Cluster state storage"
      - "Source of truth for configuration"
      - "Consensus via Raft algorithm"

  scheduler:
    role: "Assigning Pods to nodes"
    responsibilities:
      - "Evaluating constraints (resources, affinity)"
      - "Selecting the optimal node"
      - "Respecting PodDisruptionBudgets"

  controller_manager:
    role: "Control loops for desired state"
    controllers:
      - "ReplicaSet Controller"
      - "Deployment Controller"
      - "Service Controller"
      - "Node Controller"

# Worker Node components
worker_nodes:
  kubelet:
    role: "Agent on each node"
    responsibilities:
      - "Communication with Control Plane"
      - "Pod lifecycle management"
      - "Node status reporting"

  kube_proxy:
    role: "Network proxy on each node"
    responsibilities:
      - "iptables/IPVS rules for Services"
      - "Intra-cluster load balancing"

  container_runtime:
    role: "Container execution"
    options:
      - "containerd (recommended)"
      - "CRI-O"
```

이 아키텍처는 고가용성을 가능하게 합니다: Control Plane은 복제될 수 있으며, 워크로드는 Worker Node에 분산됩니다.

**Q6: 시작하지 않는 Pod를 어떻게 디버그합니까?**

Kubernetes 디버깅은 서로 다른 레이어를 분석하는 체계적인 접근 방식을 따릅니다.

```bash
# kubernetes-debugging.sh
# Workflow for debugging a failing Pod

# Step 1: Check Pod status
kubectl get pod my-app-pod -o wide
# STATUS: CrashLoopBackOff, ImagePullBackOff, Pending, etc.

# Step 2: Pod details and events
kubectl describe pod my-app-pod
# Important sections:
# - Conditions (PodScheduled, Initialized, Ready)
# - Events (scheduling, pull errors, etc.)

# Step 3: Container logs
kubectl logs my-app-pod --previous  # Previous crash logs
kubectl logs my-app-pod -c init-container  # Init container logs

# Step 4: Interactive execution for debugging
kubectl exec -it my-app-pod -- sh
# Check: env vars, mounted files, network

# Step 5: Check available resources
kubectl describe node <node-name>
# Sections: Allocatable, Allocated resources

# Step 6: Debug with ephemeral Pod (K8s 1.25+)
kubectl debug my-app-pod -it --image=busybox --share-processes
```

일반적인 원인에는 리소스 부족, 이미지를 찾을 수 없음, 누락된 시크릿, 또는 잘못 구성된 프로브가 포함됩니다.

```yaml
# pod-debugging-checklist.yaml
# Debugging checklist by status
debugging_by_status:
  Pending:
    causes:
      - "Insufficient resources on nodes"
      - "PersistentVolumeClaim not bound"
      - "Affinity/Taints not satisfied"
    commands:
      - "kubectl describe pod <name> | grep -A 20 Events"
      - "kubectl get pvc"
      - "kubectl describe nodes | grep -A 5 Allocated"

  ImagePullBackOff:
    causes:
      - "Non-existent image or incorrect tag"
      - "Private registry without imagePullSecrets"
      - "Docker Hub rate limiting"
    commands:
      - "kubectl get events --field-selector reason=Failed"
      - "kubectl get secret <pull-secret> -o yaml"

  CrashLoopBackOff:
    causes:
      - "Application error at startup"
      - "Missing configuration (env vars, configmaps)"
      - "Liveness probe too aggressive"
    commands:
      - "kubectl logs <pod> --previous"
      - "kubectl describe pod <pod> | grep -A 10 Liveness"

  OOMKilled:
    causes:
      - "Memory limit too low"
      - "Memory leak in application"
    commands:
      - "kubectl describe pod <pod> | grep -A 5 Last State"
      - "kubectl top pod <pod>"
```

## Infrastructure as Code

IaC 질문은 프로비저닝 도구와 모범 사례에 대한 숙련도를 평가합니다.

**Q7: Terraform vs Ansible: 각 도구를 언제 사용해야 합니까?**

이 두 도구는 서로 다른 철학과 사용 사례를 가지고 있습니다.

```hcl
# terraform-example.tf
# Terraform: Infrastructure provisioning (declarative)
# Ideal for: cloud resources, networking, infrastructure state

terraform {
  required_providers {
    aws = {
      source  = "hashicorp/aws"
      version = "~> 5.0"
    }
  }
  # Remote state for collaboration
  backend "s3" {
    bucket         = "terraform-state-prod"
    key            = "infrastructure/terraform.tfstate"
    region         = "eu-west-1"
    dynamodb_table = "terraform-locks"
    encrypt        = true
  }
}

# Declarative resource: Terraform manages the lifecycle
resource "aws_eks_cluster" "main" {
  name     = "production-cluster"
  role_arn = aws_iam_role.eks_cluster.arn
  version  = "1.29"

  vpc_config {
    subnet_ids              = module.vpc.private_subnets
    endpoint_private_access = true
    endpoint_public_access  = false
  }

  # Implicit dependencies managed by Terraform
  depends_on = [
    aws_iam_role_policy_attachment.eks_cluster_policy
  ]
}

# Reusable modules for standardization
module "vpc" {
  source  = "terraform-aws-modules/vpc/aws"
  version = "5.0.0"

  name = "production-vpc"
  cidr = "10.0.0.0/16"

  azs             = ["eu-west-1a", "eu-west-1b", "eu-west-1c"]
  private_subnets = ["10.0.1.0/24", "10.0.2.0/24", "10.0.3.0/24"]
  public_subnets  = ["10.0.101.0/24", "10.0.102.0/24", "10.0.103.0/24"]

  enable_nat_gateway = true
  single_nat_gateway = false  # HA: one NAT per AZ
}
```

```yaml
# ansible-example.yml
# Ansible: Configuration management (procedural/declarative)
# Ideal for: OS configuration, app deployment, orchestration

---
- name: Configure application servers
  hosts: app_servers
  become: yes
  vars:
    app_version: "2.5.0"

  tasks:
    # System package management
    - name: Install required packages
      ansible.builtin.apt:
        name:
          - nginx
          - python3-pip
          - supervisor
        state: present
        update_cache: yes

    # Configuration via Jinja2 templates
    - name: Deploy nginx configuration
      ansible.builtin.template:
        src: templates/nginx.conf.j2
        dest: /etc/nginx/sites-available/app
        owner: root
        group: root
        mode: '0644'
      notify: Reload nginx

    # Application deployment
    - name: Deploy application
      ansible.builtin.git:
        repo: "https://github.com/org/app.git"
        dest: /opt/app
        version: "v{{ app_version }}"
      notify: Restart application

  handlers:
    - name: Reload nginx
      ansible.builtin.service:
        name: nginx
        state: reloaded

    - name: Restart application
      ansible.builtin.supervisorctl:
        name: app
        state: restarted
```

요약하면: Terraform은 인프라(무엇이 존재하는지), Ansible은 설정(어떻게 구성되었는지)을 위해 사용합니다. 두 도구는 완전한 워크플로에서 결합되는 경우가 많습니다.

**Q8: 대규모 조직을 위한 Terraform 프로젝트를 어떻게 구조화합니까?**

환경 분리가 있는 모듈식 구조는 유지 관리와 협업을 용이하게 합니다.

```bash
# terraform-project-structure
# Recommended structure for enterprise projects

terraform-infrastructure/
├── modules/                    # Reusable modules
│   ├── networking/
│   │   ├── main.tf
│   │   ├── variables.tf
│   │   ├── outputs.tf
│   │   └── README.md
│   ├── kubernetes/
│   │   ├── main.tf
│   │   ├── variables.tf
│   │   └── outputs.tf
│   └── database/
│       ├── main.tf
│       ├── variables.tf
│       └── outputs.tf
│
├── environments/               # Per-environment configuration
│   ├── dev/
│   │   ├── main.tf            # Calls modules
│   │   ├── variables.tf
│   │   ├── terraform.tfvars   # Dev values
│   │   └── backend.tf         # Dev state
│   ├── staging/
│   │   ├── main.tf
│   │   ├── terraform.tfvars
│   │   └── backend.tf
│   └── production/
│       ├── main.tf
│       ├── terraform.tfvars
│       └── backend.tf
│
├── shared/                     # Shared resources
│   ├── iam/
│   └── dns/
│
└── .github/
    └── workflows/
        └── terraform.yml      # CI/CD pipeline
```

```hcl
# environments/production/main.tf
# Example of module usage

module "networking" {
  source = "../../modules/networking"

  environment     = "production"
  vpc_cidr        = var.vpc_cidr
  azs             = var.availability_zones
  enable_flow_logs = true
}

module "kubernetes" {
  source = "../../modules/kubernetes"

  environment    = "production"
  cluster_name   = "prod-cluster"
  vpc_id         = module.networking.vpc_id
  subnet_ids     = module.networking.private_subnet_ids
  node_groups    = var.node_groups

  # Production: HA configuration
  cluster_version           = "1.29"
  enable_cluster_autoscaler = true
}

module "database" {
  source = "../../modules/database"

  environment        = "production"
  vpc_id             = module.networking.vpc_id
  subnet_ids         = module.networking.database_subnet_ids
  instance_class     = "db.r6g.xlarge"
  multi_az           = true  # HA in production
  backup_retention   = 30
}
```

이 구조는 다음을 가능하게 합니다: 모듈 버전 관리, 환경별 변경 검토, 코드 재사용.

## 모니터링과 관찰 가능성

모니터링 질문은 관찰 가능한 시스템을 설계하는 능력을 평가합니다.

**Q9: 관찰 가능성의 세 가지 기둥은 무엇입니까?**

관찰 가능성은 시스템의 내부 상태를 이해하는 데 도움이 되는 세 가지 보완적인 데이터 유형에 의존합니다.

```yaml
# observability-pillars.yaml
# The three pillars of observability

pillars:
  metrics:
    description: "Numeric data aggregated over time"
    characteristics:
      - "Low cardinality"
      - "Efficient storage"
      - "Ideal for alerting"
    examples:
      - "request_count (counter)"
      - "response_time_seconds (histogram)"
      - "active_connections (gauge)"
    tools:
      - "Prometheus"
      - "Datadog"
      - "CloudWatch"
    use_cases:
      - "Real-time dashboards"
      - "Threshold alerts"
      - "Capacity planning"

  logs:
    description: "Timestamped text events"
    characteristics:
      - "High cardinality"
      - "Detailed context"
      - "Large storage"
    examples:
      - "Application errors"
      - "Audit events"
      - "Debug information"
    tools:
      - "Loki"
      - "Elasticsearch"
      - "CloudWatch Logs"
    use_cases:
      - "Debugging"
      - "Audit compliance"
      - "Root cause analysis"

  traces:
    description: "Request tracking across services"
    characteristics:
      - "End-to-end view"
      - "Context propagation"
      - "Bottleneck identification"
    examples:
      - "Distributed transaction"
      - "Service dependencies"
      - "Latency breakdown"
    tools:
      - "Jaeger"
      - "Tempo"
      - "AWS X-Ray"
    use_cases:
      - "Performance optimization"
      - "Service dependencies"
      - "Error propagation"
```

**Q10: 효과적인 알림을 어떻게 구성합니까?**

잘 설계된 알림은 피로를 줄이고 빠른 인시던트 대응을 가능하게 합니다.

```yaml
# prometheus-alerting-rules.yaml
# Prometheus alerting rules with best practices

groups:
  - name: application-alerts
    rules:
      # Alert on symptom, not cause
      - alert: HighErrorRate
        # Error rate > 1% over 5 minutes
        expr: |
          sum(rate(http_requests_total{status=~"5.."}[5m]))
          /
          sum(rate(http_requests_total[5m]))
          > 0.01
        for: 5m  # Avoid false positives
        labels:
          severity: critical
          team: backend
        annotations:
          summary: "High error rate detected"
          description: |
            Error rate is {{ $value | humanizePercentage }}
            for the last 5 minutes.
          runbook_url: "https://wiki.example.com/runbooks/high-error-rate"

      # Proactive alert on saturation
      - alert: DiskSpaceRunningLow
        expr: |
          (node_filesystem_avail_bytes / node_filesystem_size_bytes)
          * 100 < 20
        for: 15m
        labels:
          severity: warning
        annotations:
          summary: "Disk space below 20%"
          description: |
            Node {{ $labels.instance }} has only
            {{ $value | humanize }}% disk space remaining.

      # SLO-based alerting
      - alert: SLOBudgetBurnRate
        # Error budget consumed too quickly
        expr: |
          (
            sum(rate(http_requests_total{status=~"5.."}[1h]))
            /
            sum(rate(http_requests_total[1h]))
          ) > (1 - 0.999) * 14.4
        for: 5m
        labels:
          severity: critical
        annotations:
          summary: "SLO budget burning too fast"
          description: |
            At current error rate, monthly SLO budget will be
            exhausted in less than 2 days.
```

```yaml
# alertmanager-config.yaml
# AlertManager configuration with intelligent routing

global:
  resolve_timeout: 5m

route:
  receiver: default
  group_by: [alertname, cluster, service]
  group_wait: 30s       # Wait to group alerts
  group_interval: 5m    # Interval between grouped notifications
  repeat_interval: 4h   # Re-alert if not resolved

  routes:
    # Critical alerts: immediate notification
    - match:
        severity: critical
      receiver: pagerduty-critical
      continue: true  # Also notify Slack

    # Alerts by team
    - match:
        team: backend
      receiver: slack-backend

    - match:
        team: infrastructure
      receiver: slack-infra

receivers:
  - name: pagerduty-critical
    pagerduty_configs:
      - service_key: <pagerduty-key>
        severity: critical

  - name: slack-backend
    slack_configs:
      - channel: '#alerts-backend'
        send_resolved: true
        title: '{{ .Status | toUpper }}: {{ .CommonAnnotations.summary }}'
        text: '{{ .CommonAnnotations.description }}'
```

핵심 원칙: 원인이 아닌 증상(사용자 영향)에 따라 알림을 보내고, 런북을 포함시키며, SLO에 따라 임계값을 조정합니다.

## 보안과 컴플라이언스

보안 질문은 위험과 대책에 대한 이해를 평가합니다.

**Q11: Kubernetes 클러스터를 어떻게 보안합니까?**

Kubernetes 보안은 여러 레이어를 다룹니다: 네트워크, 인증, 워크로드, 데이터.

```yaml
# kubernetes-security-policies.yaml
# NetworkPolicy: network isolation between namespaces

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: default-deny-all
  namespace: production
spec:
  # Applied to all pods in namespace
  podSelector: {}
  policyTypes:
    - Ingress
    - Egress
  # No traffic allowed by default
  ingress: []
  egress: []
---
# Allow only necessary traffic
apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: api-network-policy
  namespace: production
spec:
  podSelector:
    matchLabels:
      app: api
  policyTypes:
    - Ingress
    - Egress
  ingress:
    # Accept only from ingress controller
    - from:
        - namespaceSelector:
            matchLabels:
              name: ingress-nginx
      ports:
        - protocol: TCP
          port: 8080
  egress:
    # Allow to database
    - to:
        - podSelector:
            matchLabels:
              app: postgres
      ports:
        - protocol: TCP
          port: 5432
    # Allow DNS
    - to:
        - namespaceSelector: {}
          podSelector:
            matchLabels:
              k8s-app: kube-dns
      ports:
        - protocol: UDP
          port: 53
```

```yaml
# pod-security-standards.yaml
# PodSecurity: workload restrictions

apiVersion: v1
kind: Namespace
metadata:
  name: production
  labels:
    # Enforce: blocks violations
    pod-security.kubernetes.io/enforce: restricted
    # Warn: warns without blocking
    pod-security.kubernetes.io/warn: restricted
    # Audit: logs violations
    pod-security.kubernetes.io/audit: restricted
---
# Pod compliant with "restricted" standards
apiVersion: v1
kind: Pod
metadata:
  name: secure-pod
  namespace: production
spec:
  securityContext:
    runAsNonRoot: true
    runAsUser: 1000
    fsGroup: 1000
    seccompProfile:
      type: RuntimeDefault
  containers:
    - name: app
      image: myapp:latest
      securityContext:
        allowPrivilegeEscalation: false
        readOnlyRootFilesystem: true
        capabilities:
          drop:
            - ALL
      resources:
        limits:
          memory: "256Mi"
          cpu: "500m"
        requests:
          memory: "128Mi"
          cpu: "250m"
      volumeMounts:
        - name: tmp
          mountPath: /tmp
  volumes:
    - name: tmp
      emptyDir: {}
```

> **심층 방어**
>
> Kubernetes 보안은 여러 레이어를 결합합니다: 인가를 위한 RBAC, 네트워크 격리를 위한 NetworkPolicies, 워크로드 제한을 위한 PodSecurity, 저장 시 시크릿 암호화.

**Q12: 최소 권한 원칙이란 무엇이며 어떻게 적용합니까?**

이 원칙은 사용자나 시스템이 작업을 완료하는 데 필요한 최소한의 권한만 가져야 한다고 규정합니다.

```yaml
# rbac-least-privilege.yaml
# Kubernetes RBAC with minimal permissions

# Role: permissions in a specific namespace
apiVersion: rbac.authorization.k8s.io/v1
kind: Role
metadata:
  namespace: production
  name: deployment-manager
rules:
  # Pod reading (for monitoring)
  - apiGroups: [""]
    resources: ["pods", "pods/log"]
    verbs: ["get", "list", "watch"]
  # Deployment management only
  - apiGroups: ["apps"]
    resources: ["deployments"]
    verbs: ["get", "list", "watch", "update", "patch"]
  # No create/delete on deployments
  # No access to secrets or sensitive configmaps
---
# RoleBinding: Role <-> ServiceAccount association
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
  name: deployment-manager-binding
  namespace: production
subjects:
  - kind: ServiceAccount
    name: ci-cd-deployer
    namespace: production
roleRef:
  kind: Role
  name: deployment-manager
  apiGroup: rbac.authorization.k8s.io
---
# Dedicated ServiceAccount for CI/CD
apiVersion: v1
kind: ServiceAccount
metadata:
  name: ci-cd-deployer
  namespace: production
  annotations:
    # Automatic token expiration
    kubernetes.io/enforce-mountable-secrets: "true"
```

이 원칙은 AWS IAM, 데이터베이스, 네트워크 접근에도 적용됩니다.

## SRE와 신뢰성

SRE 질문은 신뢰성 실천과 인시던트 관리에 대한 이해를 평가합니다.

**Q13: SLO란 무엇이며 어떻게 정의합니까?**

Service Level Objectives(SLO)는 서비스의 기대 신뢰성을 정량화하고 엔지니어링 결정을 안내합니다.

```yaml
# slo-definitions.yaml
# SLO definitions for an API service

service: payment-api
owner: payments-team

slos:
  - name: availability
    description: "Service responds successfully to requests"
    sli:
      # SLI: measured metric
      type: availability
      good_events: "http_requests_total{status=~'2..'}"
      total_events: "http_requests_total"
    target: 99.9%  # SLO: objective
    window: 30d    # Measurement window
    # Error budget: 0.1% = 43.2 minutes/month
    error_budget:
      monthly_minutes: 43.2

  - name: latency
    description: "Response time below threshold"
    sli:
      type: latency
      good_events: "http_request_duration_seconds_bucket{le='0.3'}"
      total_events: "http_request_duration_seconds_count"
    target: 99%  # 99% of requests < 300ms
    window: 30d

  - name: throughput
    description: "Ability to process transactions"
    sli:
      type: throughput
      query: "sum(rate(transactions_processed_total[5m]))"
    target: ">= 1000 TPS"

# Actions based on error budget
error_budget_policy:
  - condition: "remaining > 50%"
    actions:
      - "Feature development prioritized"
      - "Experimentation allowed"
  - condition: "remaining 20-50%"
    actions:
      - "Balance features and reliability"
      - "Increase testing coverage"
  - condition: "remaining < 20%"
    actions:
      - "Freeze non-critical deployments"
      - "Focus on reliability improvements"
  - condition: "exhausted"
    actions:
      - "Incident response mode"
      - "All hands on reliability"
```

SLO는 객관적인 결정을 가능하게 합니다: 새 기능 배포 vs 신뢰성 강화.

**Q14: 효과적인 포스트모템을 어떻게 진행합니까?**

블레임리스 포스트모템은 학습을 촉진하고 미래 인시던트를 예방합니다.

```yaml
# postmortem-template.yaml
# Blameless post-mortem template

incident:
  id: "INC-2026-0042"
  title: "Payment service unavailability"
  severity: SEV1
  duration: "45 minutes"
  date: "2026-01-15"

# Factual timeline
timeline:
  - time: "14:32"
    event: "Alert: error rate > 5% on payment-api"
    actor: "PagerDuty"
  - time: "14:35"
    event: "Incident declared, team notified"
    actor: "On-call engineer"
  - time: "14:42"
    event: "Cause identified: connection pool exhausted"
    actor: "Backend team"
  - time: "14:55"
    event: "Mitigation: deployment rollback"
    actor: "Backend team"
  - time: "15:17"
    event: "Service restored, monitoring stable"
    actor: "Backend team"

# Measurable impact
impact:
  users_affected: 12500
  transactions_failed: 847
  revenue_impact: "~$16,500"
  slo_budget_consumed: "2.3 days"

# Root cause analysis (5 Whys)
root_cause_analysis:
  - question: "Why was the service unavailable?"
    answer: "DB connections were exhausted"
  - question: "Why were connections exhausted?"
    answer: "A slow query was blocking connections"
  - question: "Why was there a slow query?"
    answer: "Missing index on a new table"
  - question: "Why was the index missing?"
    answer: "Incomplete migration deployed"
  - question: "Why was the migration incomplete?"
    answer: "No execution plan validation in staging"

# Corrective actions
action_items:
  - id: "AI-001"
    type: "prevent"
    description: "Add SQL execution plan validation in CI"
    owner: "DBA team"
    due_date: "2026-01-22"
    priority: P1

  - id: "AI-002"
    type: "detect"
    description: "Alert on connection pool usage > 80%"
    owner: "SRE team"
    due_date: "2026-01-18"
    priority: P1

  - id: "AI-003"
    type: "mitigate"
    description: "Implement circuit breaker on DB queries"
    owner: "Backend team"
    due_date: "2026-01-29"
    priority: P2

# Lessons learned
lessons_learned:
  what_went_well:
    - "Fast detection thanks to alerting (< 3 min)"
    - "Clear communication in incident channel"
    - "Rollback completed in less than 15 minutes"
  what_went_poorly:
    - "No load testing on new endpoint"
    - "Staging didn't reflect prod data volume"
  lucky:
    - "Incident during daytime with full team available"
```

목표는 시스템을 개선하는 것이지, 비난할 사람을 찾는 것이 아닙니다. 조치는 세 가지 범주로 분류됩니다: 예방, 감지, 완화.

## 결론

DevOps 인터뷰는 문화부터 기술 도구까지 광범위한 기술을 다룹니다. 성공의 열쇠는 구체적인 구현 예시로 설명된 개념에 대한 깊은 이해를 보여주는 데 있습니다.

### 준비 체크리스트

- ✅ CI/CD 개념을 숙달하고 완전한 파이프라인을 설계할 수 있다
- ✅ Kubernetes 아키텍처를 이해하고 일반적인 문제를 디버그할 수 있다
- ✅ IaC 도구(Terraform, Ansible)와 각각의 사용 사례를 알고 있다
- ✅ 모니터링 설정 및 관련 알림 정의 방법을 알고 있다
- ✅ 보안 모범 사례(최소 권한, 심층 방어)를 적용할 수 있다
- ✅ SRE 실천(SLO, 에러 버짓, 포스트모템)을 설명할 수 있다
- ✅ 구체적인 문제 해결 예시를 가지고 있다
- ✅ 복잡한 개념을 간단하게 설명할 수 있다

---

Source: SharpSkill (https://sharpskill.dev), tech interview preparation for your real stack.
HTML version of this page: https://sharpskill.dev/ko/blog/devops/essential-devops-interview-questions