Recovery Objectives

Measured from DR tests performed in Q1 2026.

Tier	Scope	RTO	RPO
Tier 1 — Full outage	All services down; database inaccessible	4 hours	1 hour
Tier 2 — Single-service failure	One microservice crashed; DB healthy	15 minutes	0 (stateless restart)
Tier 3 — Data corruption	Data loss in PostgreSQL or Redis	2 hours	1 hour
Tier 4 — Region loss	Entire cloud region unavailable	8 hours	1 hour

PostgreSQL WAL Archiving

WAL segments are continuously streamed to object storage using wal-g or pgBackRest, achieving RPO ≈ 1 minute in steady state. A full base backup runs daily at 02:00 UTC via a Kubernetes CronJob, retained for 30 days.

yaml

Helm values — postgresql.walArchiving

1postgresql:
2  walArchiving:
3    enabled: true
4    walg:
5      enabled: true
6      s3Bucket: "your-org-aegissovereign-wal"
7      s3Region: "us-east-1"
8    archiveTimeout: 60   # seconds

Point-in-Time Recovery (PITR)

Restore to any point within the 30-day retention window.

bash

PITR runbook

1# 1. Stop writes
2kubectl scale deployment aegissovereign-backend --replicas=0
3
4# 2. Restore base backup
5wal-g backup-fetch /var/lib/postgresql/data LATEST
6
7# 3. Set recovery target
8cat >> /var/lib/postgresql/data/postgresql.conf <<EOF
9restore_command = 'wal-g wal-fetch %f %p'
10recovery_target_time = '2026-04-10 14:30:00+00'
11recovery_target_action = 'promote'
12EOF
13touch /var/lib/postgresql/data/recovery.signal
14
15# 4. Restart
16kubectl scale deployment aegissovereign-postgres --replicas=1
17kubectl scale deployment aegissovereign-backend --replicas=3

Kubernetes Backup (Velero)

Velero backs up K8s resources, ConfigMaps, Secrets, and PVC snapshots on an hourly schedule with 7-day retention, and daily with 30-day retention.

bash

Velero schedule

1velero schedule create aegissovereign-hourly \
2  --schedule="0 * * * *" \
3  --include-namespaces aegissovereign \
4  --ttl 168h
5
6velero schedule create aegissovereign-daily \
7  --schedule="0 3 * * *" \
8  --include-namespaces aegissovereign \
9  --ttl 720h

DR Test History

DR tests are scheduled quarterly. Results are reviewed by Engineering and shared with Enterprise customers on request.

Date	Type	RTO Achieved	RPO Achieved	Notes
2026-01-15	PITR drill (Tier 3)	1h 45m	3 minutes	Restored to 48h prior timestamp
2026-02-12	Full cluster restore (Tier 1)	3h 20m	58 minutes	Within RTO/RPO targets
2026-03-05	Redis loss simulation (Tier 2)	8 minutes	0	Cache warmed from DB queries

Backup Encryption

All backups are encrypted at rest.

Layer	Encryption
WAL archives (S3/GCS)	AES-256 (SSE-S3 or CMEK)
Base backups	wal-g built-in AES-256-CTR
Velero PVC snapshots	Cloud-native volume encryption (EBS/Persistent Disk)
Redis RDB/AOF	PVC-level encryption via storage class