Web Components Disaster Recovery Plan —

DR Plan

Web Components Disaster Recovery Plan Backup RTO RPO Failover Architecture Recovery Testing 3-2-1 Rule Replication Off-site Chaos Engineering Production

เนื้อหาเกี่ยวข้อง — Delta Lake Cost Optimization ลดค่าใช้จ่าย

Backup Strategy

# === Backup Configuration ===



# PostgreSQL WAL Archiving

# postgresql.conf:

# wal_level = replica

# archive_mode = on

# archive_command = 'aws s3 cp %p s3://db-backup/wal/%f'

# max_wal_senders = 5



# Streaming Replication Setup

# primary: pg_hba.conf

# host replication replicator 10.0.0.0/8 md5

#

# replica:

# pg_basebackup -h primary -D /var/lib/postgresql/data -U replicator -P -R



# Daily Full Backup

# #!/bin/bash

# DATE=$(date +%Y%m%d)

# pg_dump -Fc mydb > /backup/mydb_.dump

# aws s3 cp /backup/mydb_.dump s3://db-backup/daily/

# find /backup -name "*.dump" -mtime +30 -delete



# File Backup with rsync

# rsync -avz --delete /var/www/html/ backup-server:/backup/www/

# aws s3 sync /var/www/uploads s3://file-backup/uploads/ --delete



# Config Backup

# git add -A && git commit -m "config backup $(date)" && git push



from dataclasses import dataclass



@dataclass

class BackupJob:

    component: str

    method: str

    frequency: str

    retention: str

    destination: str

    encryption: str

    test_frequency: str



jobs = [

    BackupJob("PostgreSQL Full", "pg_dump", "Daily 02:00", "30 days", "S3 ap-southeast-1", "AES-256", "Monthly"),

    BackupJob("PostgreSQL WAL", "WAL Archiving", "Continuous", "7 days", "S3 ap-southeast-1", "AES-256", "Monthly"),

    BackupJob("PostgreSQL Replica", "Streaming", "Real-time", "N/A", "DR Site Tokyo", "TLS", "Quarterly"),

    BackupJob("Uploads/Media", "S3 Sync", "Hourly", "90 days", "S3 cross-region", "SSE-S3", "Monthly"),

    BackupJob("Config/IaC", "Git Push", "On change", "Unlimited", "GitHub Private", "SSH", "On deploy"),

    BackupJob("Secrets", "Vault Snapshot", "Daily", "30 days", "S3 encrypted", "AES-256", "Monthly"),

]



print("=== Backup Jobs ===")

for j in jobs:

    print(f"  [{j.component}] Method: {j.method}")

    print(f"    Freq: {j.frequency} | Retention: {j.retention}")

    print(f"    Dest: {j.destination} | Encrypt: {j.encryption}")

    print(f"    Test: {j.test_frequency}")

Failover Architecture

# === Multi-region Failover ===



# DNS Failover (Route53 / Cloudflare)

# Primary: ap-southeast-1 (Singapore) — Active

# DR: ap-northeast-1 (Tokyo) — Standby

# Health Check: HTTP /health every 30s

# Failover: Auto switch DNS when primary unhealthy



# Kubernetes Multi-cluster

# kubectl config use-context primary-cluster

# kubectl get pods  # Running workloads

#

# kubectl config use-context dr-cluster

# kubectl get pods  # Standby workloads (scaled to 0 or warm)

#

# Failover script:

# kubectl --context dr-cluster scale deploy/web --replicas=3

# kubectl --context dr-cluster scale deploy/api --replicas=3

# # Update DNS to DR cluster IP



# Database Failover

# On DR site:

# pg_ctl promote -D /var/lib/postgresql/data

# # Replica becomes new primary



@dataclass

class FailoverStep:

    step: int

    action: str

    responsible: str

    time_estimate: str

    automated: bool

    verification: str



steps = [

    FailoverStep(1, "Detect failure — monitoring alert", "Monitoring System", "1-2 min", True, "Alert received in Opsgenie"),

    FailoverStep(2, "Assess scope — determine affected components", "On-call Engineer", "2-5 min", False, "Impact assessment complete"),

    FailoverStep(3, "Activate DR — switch DNS to DR site", "On-call Engineer", "1-2 min", True, "DNS propagated"),

    FailoverStep(4, "Promote DB replica", "DBA / Script", "2-5 min", True, "DB accepting writes"),

    FailoverStep(5, "Scale DR workloads", "DevOps / Script", "2-3 min", True, "All pods running"),

    FailoverStep(6, "Verify services — health checks pass", "On-call Engineer", "3-5 min", True, "All health checks green"),

    FailoverStep(7, "Notify stakeholders", "Incident Commander", "5 min", True, "Status page updated"),

    FailoverStep(8, "Monitor DR site", "Team", "Ongoing", False, "Metrics stable 30 min"),

]



print("=== Failover Runbook ===")

total_time = 0

for s in steps:

    auto = "Auto" if s.automated else "Manual"

    print(f"  Step {s.step}: {s.action}")

    print(f"    Who: {s.responsible} | Time: {s.time_estimate} | {auto}")

    print(f"    Verify: {s.verification}")

DR Testing

# === DR Test Plan ===



@dataclass

class DRTest:

    test_type: str

    frequency: str

    scope: str

    duration: str

    participants: str

    success_criteria: str



tests = [

    DRTest("Tabletop Exercise", "Monthly", "Discuss scenarios, review runbook", "1 hour",

        "All engineers + management", "Everyone knows their role"),

    DRTest("Backup Restore", "Monthly", "Restore DB from latest backup", "30 min",

        "DBA + DevOps", "DB restored, data verified"),

    DRTest("Partial Failover", "Quarterly", "Failover one component to DR", "2 hours",

        "DevOps + Backend", "Component running on DR, traffic flowing"),

    DRTest("Full Failover", "Bi-annually", "Complete switch to DR site", "4 hours",

        "All teams", "All services running on DR, RTO met"),

    DRTest("Chaos Engineering", "Monthly", "Kill random pods/nodes", "1 hour",

        "DevOps + SRE", "System self-heals, no customer impact"),

]



print("DR Test Plan:")

for t in tests:

    print(f"  [{t.test_type}] {t.frequency}")

    print(f"    Scope: {t.scope}")

    print(f"    Duration: {t.duration} | Team: {t.participants}")

    print(f"    Pass: {t.success_criteria}")



dr_metrics = {

    "Last Full DR Test": "2024-10-15 — Passed (RTO: 18 min)",

    "Last Backup Restore": "2024-11-01 — Passed (12 min)",

    "Backup Success Rate": "99.8% (2 failures in 365 days)",

    "Actual RTO (last incident)": "22 minutes",

    "Actual RPO (last incident)": "3 minutes",

    "DR Site Readiness": "Warm standby, auto-scale ready",

    "Runbook Last Updated": "2024-11-15",

}



print(f"\n\nDR Metrics:")

for k, v in dr_metrics.items():

    print(f"  [{k}]: {v}")

เคล็ดลับ

• 3-2-1: ใช้ 3-2-1 Rule สำหรับ Backup เสมอ
• Test: ทดสอบ DR Plan ทุก Quarter ไม่ใช่แค่เขียนแผน
• Automate: Automate Failover Steps ให้มากที่สุด
• Runbook: เขียน Runbook ชัดเจน ทุกคนทำตามได้
• Monitor: Monitor Backup Status ทุกวัน Alert เมื่อ Failed

Disaster Recovery Plan คืออะไร

แผนกู้คืนระบบ Server ล่ม ไฟไหม้ Ransomware RTO RPO ขั้นตอนกู้คืน ทดสอบ Quarter อัปเดตเมื่อเปลี่ยน Business Impact

แนะนำเพิ่มเติม — เรียนเทรดกับ iCafeForex

เนื้อหาเกี่ยวข้อง — อ่านต่อ: LocalAI Self-hosted Code Review Best Practice — คู่มือฉบับสมบูรณ์ 2026

เนื้อหาเกี่ยวข้อง — บทความที่เกี่ยวข้อง: Prometheus Alertmanager Monitoring และ Alerting