---
last_verified: 2026-03-14
pattern: Database Disaster Recovery
category: database
---

description: |
  DR patterns for Oracle databases on OCI. Covers active-passive,
  active-active, and pilot light approaches.

tiers:
  tier_1_mission_critical:
    name: "Mission Critical"
    rto: "<15 minutes"
    rpo: "<1 minute"
    pattern: "Active-passive with automatic failover"
    implementation:
      adb_s: "Cross-region ADG with auto-failover"
      exacs: "Cross-region Data Guard with FSFO + Observer"
    cost_multiplier: "2x (full standby)"
    testing: "Quarterly automated DR drill"

  tier_2_business_critical:
    name: "Business Critical"
    rto: "<1 hour"
    rpo: "<15 minutes"
    pattern: "Active-passive with manual failover"
    implementation:
      adb_s: "Cross-region ADG with manual switchover"
      exacs: "Cross-region Data Guard without FSFO"
    cost_multiplier: "2x (full standby) or 1.5x (scaled-down standby)"
    testing: "Semi-annual DR drill"

  tier_3_standard:
    name: "Standard"
    rto: "<4 hours"
    rpo: "<1 hour"
    pattern: "Pilot light or backup-restore"
    implementation:
      adb_s: "Cross-region backup with restore-on-demand"
      exacs: "RMAN backups to Object Storage, cross-region replicated"
    cost_multiplier: "1.1x (backup storage only)"
    testing: "Annual DR drill"

dr_regions:
  recommended_pairs:
    - primary: "US East (Ashburn)"
      dr: "US West (Phoenix)"
      latency_ms: "~60"
    - primary: "UK South (London)"
      dr: "Germany Central (Frankfurt)"
      latency_ms: "~15"
    - primary: "Japan East (Tokyo)"
      dr: "Japan Central (Osaka)"
      latency_ms: "~10"
    - primary: "Brazil East (Sao Paulo)"
      dr: "Brazil Southeast (Vinhedo)"
      latency_ms: "~5"

migration_cutover:
  zero_downtime:
    tools: ["ZDM with GoldenGate", "Data Guard switchover"]
    prerequisites:
      - "GoldenGate configured and replicating"
      - "Application connection strings use TNS aliases"
      - "DNS TTL reduced before cutover"
    steps:
      - "Verify replication lag is zero"
      - "Stop writes to source"
      - "Verify final sync"
      - "Switch application connection strings"
      - "Validate application functionality"
      - "Decommission source (after bake period)"

  planned_downtime:
    tools: ["Data Pump", "RMAN restore", "ZDM physical"]
    typical_downtime: "2-8 hours depending on database size"
    steps:
      - "Announce maintenance window"
      - "Take final backup/export"
      - "Restore/import to target"
      - "Switch application connections"
      - "Validate"

gotchas:
  - id: DR-001
    issue: "Cross-region ADG network latency affects commit performance"
    impact: "Synchronous mode adds latency to every commit"
    recommendation: "Use asynchronous for cross-region (default), accept seconds of RPO"
  - id: DR-002
    issue: "DR standby compute may not be available during regional disaster"
    impact: "If DR region is also impacted, compute capacity may be limited"
    recommendation: "Use capacity reservations in DR region for critical workloads"
  - id: DR-003
    issue: "Application tier DR often forgotten"
    impact: "Database fails over but app tier is not available"
    recommendation: "Include app tier in DR plan — Terraform for on-demand provisioning"