Ekspluatatsiya

Backup va restore

PostgreSQL, S3, encryption key va operatsion restore tartibi.

Nimani saqlash kerak?

Aktiv	Nega kerak	Talab
PostgreSQL	Barcha metadata, chat, auth, jobs, usage	Encrypted, off-site, retention + PITR
S3 bucket	Foydalanuvchi fayl baytlari	Versioning/lifecycle yoki replikatsiya
Master encryption key	Provider credentiallarini decrypt qilish	Alohida secret manager backup
Deploy config	Image digest, Caddy va env nomlari	Secret qiymatlaridan ajratilgan config-as-code
Caddy state	ACME account/cert cache	Persistent volume; qayta olish ham mumkin

Kalitsiz backup to‘liq emas

Database restore qilinib, o‘sha MASTER_ENCRYPTION_KEY tiklanmasa provider credential ciphertextlari ochilmaydi. Kalitni DB backup bilan bir joyda saqlamang.

PostgreSQL backup

Logical backup misoli

pg_dump --format=custom --no-owner --no-acl \
  --dbname="$DATABASE_URL" \
  --file=edubase-$(date +%F-%H%M).dump

pg_restore --list edubase-2026-07-25-1200.dump | head

Katta production bazada managed PostgreSQL snapshot va WAL/PITR afzal. pg_dump consistency beradi, lekin RPO/RTO talabini yakka o‘zi qondirmasligi mumkin.

Restore drill

Izolyatsiyalangan yangi PostgreSQL/S3 target yarating.
Database va mos object versiyalarini bir vaqt nuqtasiga tiklang.
Master key va kerakli runtime secretlarni secret managerdan ulang.
Amaldagi binaryga mos migration statusini tekshiring.
API/worker’ni private staging’da ishga tushiring.
Login, provider decrypt/test, conversation, download va usage reconciliationni tekshiring.
RPO/RTO va topilgan muammolarni drill jurnaliga yozing.

DB va S3 consistency

DB metadata va S3 bytes bitta distributed transactionda emas. Restore vaqt nuqtalari farq qilsa orphan object yoki missing object paydo bo‘lishi mumkin. Periodik reconciliation va delete cleanup joblari bunday farqlarni aniqlash/tuzatishga xizmat qiladi.