Disaster Recovery
Intermediate 14 min read Disaster Recovery

Backup Strategies

The golden rule: backups you can't restore are art

Backups are the DR safety net: snapshots, point-in-time recovery, cross-region copies, immutability (ransomware/accident), retention policies, and restore verification. The discipline: test restores, not just backups.

backupsnapshotPITRretentionimmutabilityrestorecross-region
Prerequisites: DR Fundamentals Object Storage
Before we start

What you will learn

  • Choose snapshot vs PITR vs database-native backup
  • Design retention + cross-region + immutability
  • Automate restore verification
In 60 seconds

Backup types: VM/EBS snapshots, database PITR (WAL/binlog replay to a second), managed DB automated backups, object-storage copies/versioning. Strategy = schedule + retention tiers + cross-region copy + immutable vault. Proof = regular restore drills.

Introduction

What is this?

Tools: AWS Backup (vaults, VTLs, cross-region copy, policies), EBS snapshots, RDS automated backups + PITR, Aurora, Azure Backup (Recovery Services vault, MARS agent, LRS/GRS/GZRS), GCP Cloud Backup and DR, AlloyDB/Cloud SQL backups, GCS versioning/Object holds. Concepts: retention (short daily, longer weekly/monthly, archive), cold tiers for cost, immutability (Object Lock / write-once) against ransomware and fat-finger deletes, restoration verification (recover to isolate and verify data), backup testing as part of CI/DR. 3-2-1 rule: 3 copies, 2 media, 1 offsite. Pitfalls: backing up to same region, relying on the provider backup that is delete-on-close, restore permissions missing.

Motivation

Why do you need it?

  • Backups are the last line of defense against accidental delete, corruption, and ransomware.
  • Cross-region copies and immutability turn a data loss incident into a data retrieval.
  • Restore drills find what 'we have backups' actually means.
  • Audits across the industry hinge on 'we can restore from X' evidence.
Think about it

Mental model

Imagine

A safe-deposit box + copy in your own home: one is handy, one is off-site

Think of

3-2-1 backups

You need the copy at home (fast) and the one in the vault (catastrophe); a single copy at the bank is not a backup strategy, it's hope. Same for same-region-only snaps.

Mechanics

How it works

  • Provider snapshots app/db storage at cadence.
  • Retention policy keeps daily/weekly/monthly, archiving old to cold storage.
  • Copy vault + snapshot cross-region on a schedule (or on write).
  • Immutability protects copies for a duration.
  • Restore drills restore a known dataset to isolation and verify.

Backup lifecycle.

Workload
Vault
Retention
Cross-region
Restore drill
Hands-on

Examples

AWS Backup plan (pseudo)

AWS Backup plan (pseudo)
plan: "mcp-db-backups"
selections: [ db-arn ]
schedule: cron(0 2 * * ? *)   # daily
completion: 8
lifecycle:
  cold: 7; delete_after: 90
copy_action:
  destination_region: us-west-2
  lifecycle: { cold: 30, delete_after: 365 }
vault_lock: ON  # immutable, hard stop deletion

Daily vaulting, cold after a week, cross-region immutable copy — restore targets are unambiguous.

Real world

In production

  • Managed DB PITR + daily vaulted snapshots with cross-region copy.
  • Object storage versioning + lifecycle to cold, with write-object-lock on audit buckets.
  • Monthly restore drill: restore to staging RDS, verify row counts/checksum, then delete.
Guidelines

Best practices

  • 3-2-1: at least one copy off-site and ideally immutable.
  • PITR beats HBI for DBs where minutes of data matter.
  • Restore verification on schedule — the only honest backup test.
  • Keep backup IAM scoped and documented (restore after DR = someone can actually run it).
  • Audit backup coverage for every workload quarterly (new services often have none).
Watch out

Common mistakes

  • Snapshots in the same region only — the whole regional copy dies with the region.
  • Retention battles cost: deleting early is how the one needed copy vanishes.
  • No restore permissions/runbook: the backup exists, nobody can use it in a hurry.
  • Trusting provider defaults (e.g., 'PITR enabled' without cross-region coverage).
When things break

Debugging tips

  • Restore fails with permissions → the vault/role that can restore wasn't tested; verify in the drill.
  • Backup bill ballooning → lifecycle/retention drift; check the cold tier for stale copies.
  • Cross-region copy not appearing → region replication lag or the vault copy action misconfigured.
  • Restore gives inconsistent DB → you backed up the volume without quiescing; use app-consistent backups or DB-native PITR.
Interviews

Interview checkpoint

Test yourself

Quick quiz

0/2
1

The real proof a backup strategy works is:

2

Immutability in backups protects against:

Go deeper

Official documentation

AWS Backup docs
docs
Azure Backup docs
docs
Google Cloud Backup and DR
docs