.gitea/workflows/prod-deploy.yaml

@@ -77,7 +77,7 @@ jobs:
           # The main-stack images via compose (reuses the build args, incl. VERSION);
           # the bot separately, since it is profiled out of the prod compose.
           docker compose -f docker-compose.yml -f docker-compose.prod.yml build
-          docker compose -f docker-compose.yml -f docker-compose.prod.yml push backend gateway landing validator renderer
+          docker compose -f docker-compose.yml -f docker-compose.prod.yml push postgres backend gateway landing validator renderer
           docker build -f ../platform/telegram/Dockerfile --target bot --build-arg VERSION="$TAG" -t "$REGISTRY/scrabble-telegram-bot:$TAG" ..
           docker push "$REGISTRY/scrabble-telegram-bot:$TAG"
 
@@ -135,6 +135,18 @@ jobs:
       DICT_VERSION: ${{ vars.DICT_VERSION }}
       POSTGRES_DB: ${{ vars.PROD_POSTGRES_DB }}
       POSTGRES_USER: ${{ vars.PROD_POSTGRES_USER }}
+      # Point-in-time recovery (pgBackRest -> S3), prod main host only. Endpoint/bucket/
+      # region + the archive-mode switch are variables; the S3 keys + the repository cipher
+      # passphrase are secrets. All empty/off until the operator arms archiving; flipping
+      # PROD_PGBACKREST_ARCHIVE_MODE=on activates it (deploy/README.md, arming).
+      PGBACKREST_ARCHIVE_MODE: ${{ vars.PROD_PGBACKREST_ARCHIVE_MODE }}
+      PGBACKREST_S3_ENDPOINT: ${{ vars.PROD_PGBACKREST_S3_ENDPOINT }}
+      PGBACKREST_S3_PORT: ${{ vars.PROD_PGBACKREST_S3_PORT }}
+      PGBACKREST_S3_BUCKET: ${{ vars.PROD_PGBACKREST_S3_BUCKET }}
+      PGBACKREST_S3_REGION: ${{ vars.PROD_PGBACKREST_S3_REGION }}
+      PGBACKREST_S3_KEY: ${{ secrets.PROD_PGBACKREST_S3_KEY }}
+      PGBACKREST_S3_KEY_SECRET: ${{ secrets.PROD_PGBACKREST_S3_KEY_SECRET }}
+      PGBACKREST_CIPHER_PASS: ${{ secrets.PROD_PGBACKREST_CIPHER_PASS }}
       # TELEGRAM_MINIAPP_URL and GRAFANA_ROOT_URL are derived from PUBLIC_BASE_URL in
       # deploy/write-prod-env.sh, not stored variables.
     steps:

.gitea/workflows/prod-rollback.yaml

@@ -77,6 +77,16 @@ jobs:
       SERVICE_EMAIL: ${{ vars.PROD_SERVICE_EMAIL }}
       GRAFANA_SMTP_PORT: ${{ vars.GRAFANA_SMTP_PORT }}
       GF_SMTP_ENABLED: ${{ vars.PROD_GF_SMTP_ENABLED }}
+      # PITR archiving parity: a rollback must re-render the SAME env.sh, else it would
+      # silently disarm WAL archiving (PGBACKREST_ARCHIVE_MODE would fall back to off).
+      PGBACKREST_ARCHIVE_MODE: ${{ vars.PROD_PGBACKREST_ARCHIVE_MODE }}
+      PGBACKREST_S3_ENDPOINT: ${{ vars.PROD_PGBACKREST_S3_ENDPOINT }}
+      PGBACKREST_S3_PORT: ${{ vars.PROD_PGBACKREST_S3_PORT }}
+      PGBACKREST_S3_BUCKET: ${{ vars.PROD_PGBACKREST_S3_BUCKET }}
+      PGBACKREST_S3_REGION: ${{ vars.PROD_PGBACKREST_S3_REGION }}
+      PGBACKREST_S3_KEY: ${{ secrets.PROD_PGBACKREST_S3_KEY }}
+      PGBACKREST_S3_KEY_SECRET: ${{ secrets.PROD_PGBACKREST_S3_KEY_SECRET }}
+      PGBACKREST_CIPHER_PASS: ${{ secrets.PROD_PGBACKREST_CIPHER_PASS }}
       INPUT_TARGET: ${{ inputs.target_version }}
     steps:
       - uses: actions/checkout@v4

PLAN.md

@@ -33,7 +33,7 @@ status — without re-deriving decisions.
 | E1 | Trusted platform signal | 1 | DONE |
 | E2 | Currency + benefit core | 1 | DONE |
 | E3 | Wallet UI | 1 | DONE |
-| E4 | Durability (PITR) | 2 | TODO |
+| E4 | Durability (PITR) | 2 | WIP |
 | E5 | Payment intake | 2 | TODO |
 | E6 | Ads | 2 | TODO |
 | E7 | Admin & reports | 2 | TODO |
@@ -446,40 +446,57 @@ noun agreement). Svelte whitespace/`$state` naming gotchas apply.
 
 ## E4 — Durability (PITR)
 
-**Status:** TODO · **Release 2** · depends on: E0 (schema exists) · mechanics: PAYMENTS §14.
+**Status:** WIP — repo artifacts landed; **prod arming pending** (owner-coordinated, before
+E5). · **Release 2** · depends on: E0 (schema exists) · mechanics: PAYMENTS §14 (D4).
 
 **Goal.** Continuous WAL archiving with point-in-time recovery, armed **before the first
 real money** is accepted (E5 prod). Protects both money and game data.
 
-**Work.**
+**Locked decisions (this stage).**
 
-- Add WAL archiving to the prod Postgres (pgBackRest or WAL-G): base backups + continuous
+- **Tool: pgBackRest.** **Destination: Selectel S3** object storage (encrypted AES-256-CBC,
-  WAL to a second host or object storage. Wire into `deploy/` (compose/prod overlay +
+  path-style addressing), **prod main host only** — the test contour never archives. (D4 left
-  ansible `deploy/ansible/`), config + secrets under the `PROD_` prefix.
+  the tool + destination open; resolved here. Warm replica stays deferred, per D4.)
-- Restore runbook in `deploy/README.md`: base + WAL replay to a timestamp; test the restore
+- **Retention 30 days**, **daily full base backup** (a systemd timer) + continuous WAL
-  on a scratch target.
+  (`archive_command`, a 5-minute forced switch bounds the recovery point). Assessment
-- Keep the existing migration-time `pg_dump` (`deploy/prod-deploy.sh`) as belt-and-braces.
+  (owner-reviewed prod gate): measured prod WAL **~0.77 MB/day**, DB **~9.6 MB** → archive
+  **< 1 ₽/month** on Selectel S3, **negligible** perf impact. Recorded in `deploy/README.md`.
+- **Archiving ships gated OFF** (`PGBACKREST_ARCHIVE_MODE` on the DB, `pitr_enabled` for the
+  timer — both default off), so the merged/redeployed artifact is inert until armed and an
+  un-armed prod deploy **cannot** pile WAL onto the disk.
 
-**Mandatory pre-prod assessment (owner-required, gates the Release 2 prod rollout).**
+**Work (landed in the artifacts PR into `development`).**
-Before the first money goes live, produce and record here:
 
-1. **Disk-storage cost estimate** for the WAL archive + base backups (retention window ×
+- pgBackRest in the DB image (`deploy/postgres/Dockerfile`); postgres becomes a built + pushed
-   WAL volume; account for both hosts / object-storage pricing). This can change hosting
+  image (prod overlay `image:` + `prod-deploy.yaml` build/push list).
-   sizing.
+- Repository config + `archive_mode` via `PGBACKREST_*` env on the prod-overlay postgres
-2. **PG performance-degradation assessment** from continuous archiving (write amplification,
+  service; secrets/vars rendered by `deploy/write-prod-env.sh` from the `PROD_PGBACKREST_*`
-   archive_command latency, backup I/O contention) on the prod-sized instance.
+  Gitea set; **parity added to `prod-rollback.yaml`** (a rollback must not disarm archiving).
+- Daily base-backup systemd timer (Ansible `main` role, gated by `pitr_enabled`).
+- Two Grafana alerts on the exporter's `pg_stat_archiver` metrics (failing / stalled),
+  absent/NaN-safe on the contour.
+- Belt-and-braces `pg_dump` fixed to dump the **whole DB** (`deploy/prod-deploy.sh` was
+  `-n backend`, silently excluding `payments`); manual-restore runbook updated.
+- Full PITR runbook + arming sequence + recorded assessment in `deploy/README.md`.
 
-Neither blocks building the plan, but both **must** be completed and reviewed before the
+**Prod arming (SEPARATE — owner-coordinated, before E5; NOT the artifacts PR).** Owner creates
-prod release — surface the numbers to the owner (they may change host settings).
+the Selectel S3 bucket + the `PROD_PGBACKREST_*` secrets/variables (incl. `ARCHIVE_MODE=on`);
+then promote `development → master` → `prod-deploy` (archive_mode on behind the maintenance
+window), `pgbackrest stanza-create` + first base backup + `check`, re-run Ansible with
+`-e pitr_enabled=true`, and run the restore drill on an isolated one-shot target. Exact steps:
+`deploy/README.md` (point-in-time recovery — arming).
 
-**Tests.** Restore drill on a scratch instance (base + WAL → target timestamp) documented
+**Tests.** Restore drill on an isolated one-shot instance (base + WAL → target timestamp),
-and passing. No app-level tests.
+recorded in `deploy/README.md`. No app-level tests. Local verification: `docker compose config`
+valid + the custom PG image builds (archiving inert on the contour).
 
-**Done-criteria.** WAL archiving live on prod PG; a timed restore verified; cost + perf
+**Done-criteria.** Artifacts merged with archiving inert. **Fully done** at arming: WAL
-assessment recorded and owner-reviewed; runbook in `deploy/README.md`.
+archiving live on prod PG; a timed restore verified; cost + perf assessment reviewed; runbook
+current in `deploy/README.md`.
 
-**Notes/risks.** The tg host is weak (1 vCPU) — if the archive lands there, watch
+**Notes/risks.** The repository **cipher passphrase is unrecoverable if lost** — stored apart
-contention. Migrations stay expand-contract so image rollback remains DB-safe alongside PITR.
+from the S3 keys. Enabling `archive_mode` restarts postgres (rides the prod-deploy maintenance
+window). Migrations stay expand-contract so image rollback remains DB-safe alongside PITR.
 
 ---
 

deploy/.env.example

@@ -13,6 +13,22 @@ POSTGRES_DB=scrabble
 POSTGRES_USER=scrabble
 POSTGRES_PASSWORD=change-me            # required
 
+# --- Point-in-time recovery (pgBackRest -> S3; PROD main host only) ----------
+# Continuous WAL archiving for PITR. PROD-ONLY: the test contour never archives (these
+# stay unset there and archive_mode stays off). The artifact ships DISARMED —
+# PGBACKREST_ARCHIVE_MODE defaults off; arm it only after setting up the S3 repository +
+# secrets and creating the stanza, then flip it on (deploy/README.md, point-in-time
+# recovery). Gitea: PROD_PGBACKREST_* — endpoint/bucket/region/archive-mode are variables,
+# the two S3 keys + the cipher passphrase are secrets.
+PGBACKREST_ARCHIVE_MODE=off            # on = archiving active (arm only after the stanza exists)
+PGBACKREST_S3_ENDPOINT=                # S3 HOST ONLY — no https://, no port, no bucket (e.g. s3.ru-1.storage.selcloud.ru); path-style addressing
+PGBACKREST_S3_PORT=                    # optional; default 443 — set only if the provider uses a non-standard S3 port
+PGBACKREST_S3_BUCKET=                  # bucket name; lowercase, no dots/underscores
+PGBACKREST_S3_REGION=                  # e.g. ru-1
+PGBACKREST_S3_KEY=                     # secret: S3 access key id
+PGBACKREST_S3_KEY_SECRET=              # secret: S3 secret access key
+PGBACKREST_CIPHER_PASS=                # secret: repo encryption passphrase — KEEP SAFE, stored apart from the S3 keys (losing it makes the archive unrecoverable)
+
 # --- Dictionary -------------------------------------------------------------
 # scrabble-dictionary release tag baked into the image as the SEED dictionary for a
 # FRESH volume (image build-arg; also labels the resident seed version). After first

deploy/README.md

@@ -220,10 +220,89 @@ release tag, so any prior release is reachable.
 **Migrations** must be **expand-contract** (backward-compatible; goose is forward-only):
 the automatic rollback is image-only and never restores the DB. A deploy that changes
 `backend/internal/postgres/migrations/` opens a maintenance window — the backend (sole
-writer) is stopped for a consistent `pg_dump` into `/opt/scrabble/dumps` before the new
+writer) is stopped for a consistent **whole-database** `pg_dump` into `/opt/scrabble/dumps`
-backend migrates. **Manual DB restore** (only if a migration was destructive):
+(it covers `backend` **and** `payments`) before the new backend migrates. **Manual DB
-`docker exec -i scrabble-postgres psql -U scrabble -d scrabble -c 'DROP SCHEMA backend CASCADE'`,
+restore** (only if a migration was destructive): drop the app schemas in the same instance
-then pipe the dump into the same `psql`, and redeploy the matching old tag.
+and pipe the dump back —
+`docker exec -i scrabble-postgres psql -U scrabble -d scrabble -c 'DROP SCHEMA IF EXISTS backend CASCADE; DROP SCHEMA IF EXISTS payments CASCADE;'`,
+then `docker exec -i scrabble-postgres psql -U scrabble -d scrabble < the-dump.sql`, and
+redeploy the matching old tag. This dump is a belt-and-braces net for a bad migration;
+**point-in-time recovery** (below) is the primary recovery path once armed, and the only one
+that survives losing the host.
+
+## Point-in-time recovery (PITR)
+
+The main host archives Postgres continuously with **pgBackRest** to **Selectel S3**
+(encrypted at rest, path-style addressing) so the database can be restored to any moment —
+protecting the money ledger and the game state against corruption or host loss. It is the
+primary recovery path; the migration-window `pg_dump` above is the secondary net.
+
+**Shape.** A daily full **base backup** (a systemd timer on the main host, `04:00`) plus
+**continuous WAL** archived by Postgres `archive_command` (a segment is forced at least every
+5 minutes, so the recovery point is never more than a few minutes behind). Retention is **30
+days** (`repo1-retention-full=30`); the repository is AES-256-CBC encrypted. This is main-host
+only — the test contour never archives.
+
+**Wiring.** pgBackRest ships inside the DB image (`deploy/postgres/Dockerfile`); the
+repository + S3 credentials + cipher are the `PGBACKREST_*` environment on the postgres
+service in `docker-compose.prod.yml`, rendered from the `PROD_` Gitea set by
+`write-prod-env.sh`. Archiving is gated by **`PGBACKREST_ARCHIVE_MODE` (default `off`)**, so
+shipping or redeploying this stack does **not** start archiving — the artifact is inert until
+armed, which is why an un-armed prod deploy can never pile WAL onto the disk. The base-backup
+timer is provisioned by the Ansible `main` role behind `pitr_enabled` (also default off). Two
+Grafana alerts watch health: `WAL archiving failing` (`pg_stat_archiver_failed_count` rising)
+and `WAL archiving stalled` (`pg_stat_archiver_last_archive_age` over 30 min) — both
+absent/NaN-safe, so they stay quiet until archiving is armed.
+
+**Assessment (owner-reviewed; the gate before the first real money).** Measured on prod
+`pg_stat_wal`: WAL is generated at **~0.77 MB/day** and the database is **~9.6 MB**. At
+30-day retention on Selectel S3 (~2 ₽/GB·month) the archive is **under ~0.3 GB → well under
+1 ₽/month** (compression halves it again); request volume is trivial. Performance impact is
+**negligible**: archive-push moves tiny compressed segments, and the daily full base backup
+is a ~10 MB, sub-second job on the 2 vCPU host. Revisit both if traffic grows ~100× (watch
+`node_exporter` during a base backup).
+
+**Arming (owner-coordinated, once, before real payments).** Ships disarmed; to turn it on:
+
+1. **Owner (Selectel + Gitea):** create an S3 bucket (name **lowercase, no dots/underscores**)
+   and an S3 access key/secret; pick a repository **cipher passphrase** and store it **apart
+   from the S3 keys** (losing it makes the archive unrecoverable). Set the Gitea `PROD_` set —
+   variables `PROD_PGBACKREST_S3_ENDPOINT` (the S3 **host only** — no `https://`, no port, no
+   bucket) / `_S3_BUCKET` / `_S3_REGION`, an optional `_S3_PORT` (default 443, set only for a
+   non-standard provider port), and `PROD_PGBACKREST_ARCHIVE_MODE=on`; secrets
+   `PROD_PGBACKREST_S3_KEY` (the S3 **access key**) / `_S3_KEY_SECRET` (its paired **secret
+   key**, shown once at creation) / `_CIPHER_PASS` (a fresh repository passphrase, e.g.
+   `openssl rand -base64 48`).
+2. Promote `development → master`, tag, and run **prod-deploy**. The roll recreates postgres
+   with `archive_mode=on` behind the maintenance page. (Archive pushes fail harmlessly for the
+   minute until step 3 creates the repository — the WAL is retained, not lost.)
+3. On the main host, create the repository, take the first base backup, and verify:
+   ```sh
+   docker exec scrabble-postgres pgbackrest --stanza=scrabble stanza-create
+   docker exec scrabble-postgres pgbackrest --stanza=scrabble --type=full backup
+   docker exec scrabble-postgres pgbackrest --stanza=scrabble check
+   ```
+4. Enable the daily timer: `ansible-playbook site.yml -e pitr_enabled=true` (installs + starts
+   `pgbackrest-backup.timer` on the main host).
+5. Confirm in Grafana that the two archiving alerts are green (`last_archive_age` now tracks a
+   real number, `failed_count` flat).
+
+**Restore drill (proves recoverability; re-run after arming and after any major change).** On
+an **isolated, one-shot** target (a throwaway VM or a container with no ingress), pgBackRest,
+the matching Postgres major, an empty `PGDATA`, and the same `PGBACKREST_*` environment:
+
+```sh
+pgbackrest --stanza=scrabble --type=time "--target=YYYY-MM-DD HH:MM:SS+00" --delta restore
+# start Postgres; it replays WAL to the target; then verify a known row / the ledger tail
+```
+
+The target holds **real money + personal data** while it exists — keep it network-isolated and
+**destroy it (wipe `PGDATA` + the instance) afterwards**. Record each drill (date, target
+timestamp, outcome) here:
+
+| Date | Target timestamp | Result |
+| --- | --- | --- |
+| _pending first arming_ | — | — |
 
 **bot-link cert rotation:** regenerate (`deploy/gen-certs.sh /tmp/c --force`), reset the
 five `PROD_BOTLINK_*` secrets from `/tmp/c`, and re-run the workflow — both hosts redeploy
@@ -249,14 +328,17 @@ VITE_VK_APP_LINK, VITE_VK_APP_ID`. **Derived from `PUBLIC_BASE_URL` at deploy**
 `PROD_{POSTGRES_PASSWORD, GM_BASICAUTH_HASH, GRAFANA_ADMIN_PASSWORD, TELEGRAM_BOT_TOKEN,
 TELEGRAM_PROMO_BOT_TOKEN, EXPORT_SIGN_KEY, GATEWAY_HONEYTOKEN, REGISTRY_PASSWORD, SSH_KEY,
 SSH_KNOWN_HOSTS, BOTLINK_CA, BOTLINK_GATEWAY_CERT, BOTLINK_GATEWAY_KEY, BOTLINK_BOT_CERT,
-BOTLINK_BOT_KEY}`; variables:
+BOTLINK_BOT_KEY, PGBACKREST_S3_KEY, PGBACKREST_S3_KEY_SECRET, PGBACKREST_CIPHER_PASS}`;
+variables:
 `PROD_{REGISTRY_USER, MAIN_HOST, TG_HOST, CADDY_SITE_ADDRESS, GM_BASICAUTH_USER, LOG_LEVEL,
 TELEGRAM_GAME_CHANNEL_ID, TELEGRAM_CHAT_ID, TELEGRAM_SUPPORT_CHAT_ID, TELEGRAM_BOT_USERNAME,
 VITE_TELEGRAM_BOT_ID, VITE_TELEGRAM_LINK, VITE_TELEGRAM_GAME_CHANNEL_NAME, SMTP_RELAY_FROM,
 PUBLIC_BASE_URL, SMTP_RELAY_ADMIN_FROM, ADMIN_EMAIL, SMTP_RELAY_SERVICE_FROM, SERVICE_EMAIL,
-GF_SMTP_ENABLED}`. The test contour uses the same names under `TEST_`, minus the prod-only
+GF_SMTP_ENABLED, PGBACKREST_S3_ENDPOINT, PGBACKREST_S3_PORT, PGBACKREST_S3_BUCKET,
-infra (`MAIN_HOST`/`TG_HOST`/`REGISTRY_*`/`SSH_*`/`BOTLINK_*`, which the test deploy generates
+PGBACKREST_S3_REGION, PGBACKREST_ARCHIVE_MODE}`. The test contour uses the same names under `TEST_`, minus the
-or runs locally) and plus `TEST_AWG_CONF` (the bot's VPN egress).
+prod-only infra (`MAIN_HOST`/`TG_HOST`/`REGISTRY_*`/`SSH_*`/`BOTLINK_*` and the `PGBACKREST_*`
+PITR set, which is prod-only — the test contour never archives) and plus `TEST_AWG_CONF` (the
+bot's VPN egress).
 
 ## Host-side setup (outside this repo)
 

deploy/ansible/README.md

@@ -45,5 +45,7 @@ safe (idempotent) and survives a host resize.
   10m×3 log rotation), `deploy` user (docker group, no sudo), key-only sshd,
   `ufw` default-deny incoming + allow SSH, fail2ban sshd jail, unattended
   upgrades, chrony, `/opt/scrabble/{config,certs,dumps,images}`.
-- **main**: `ufw` opens 80/443/9443; the external `edge` docker network.
+- **main**: `ufw` opens 80/443/9443; the external `edge` docker network; the pgBackRest
+  daily base-backup systemd timer — installed only when `pitr_enabled=true` (off by default;
+  turned on as part of arming point-in-time recovery, see [`../README.md`](../README.md)).
 - **tg**: verifies direct `api.telegram.org` egress (the no-VPN assumption).

deploy/ansible/group_vars/all.yml

@@ -27,3 +27,12 @@ docker_log_max_file: "3"
 # path (a slow service beats a killed one). Set swap_size to "0" to skip provisioning.
 swap_size: "1G"
 swap_swappiness: 10
+
+# Point-in-time recovery (pgBackRest). The base-backup systemd timer on the main host is
+# gated by pitr_enabled so provisioning stays inert until archiving is armed — the timer
+# would fail nightly if it ran before the S3 repository + stanza exist. Arm it as part of
+# the PITR rollout (deploy/README.md): flip it on with `-e pitr_enabled=true` (or a main
+# host_var) once the stanza is created. Only the `main` role reads these.
+pitr_enabled: false
+# systemd OnCalendar for the daily full base backup — a low-traffic hour on the main host.
+pitr_backup_oncalendar: "*-*-* 04:00:00"

deploy/ansible/roles/main/tasks/main.yml

@@ -16,3 +16,30 @@
   community.docker.docker_network:
     name: edge
     state: present
+
+# --- pgBackRest base-backup schedule (point-in-time recovery) ------------------
+# A daily full base backup via a systemd timer that runs pgBackRest inside the postgres
+# container; archive_command handles the continuous WAL between backups. Gated by
+# pitr_enabled so the schedule stays inert until archiving is armed (the S3 repository +
+# stanza must exist first, or the run fails nightly). Arm per deploy/README.md.
+- name: Install the pgBackRest base-backup service
+  ansible.builtin.template:
+    src: pgbackrest-backup.service.j2
+    dest: /etc/systemd/system/pgbackrest-backup.service
+    mode: "0644"
+  when: pitr_enabled | default(false) | bool
+
+- name: Install the pgBackRest base-backup timer
+  ansible.builtin.template:
+    src: pgbackrest-backup.timer.j2
+    dest: /etc/systemd/system/pgbackrest-backup.timer
+    mode: "0644"
+  when: pitr_enabled | default(false) | bool
+
+- name: Enable and start the pgBackRest base-backup timer
+  ansible.builtin.systemd:
+    name: pgbackrest-backup.timer
+    enabled: true
+    state: started
+    daemon_reload: true
+  when: pitr_enabled | default(false) | bool

deploy/ansible/roles/main/templates/pgbackrest-backup.service.j2

@@ -0,0 +1,15 @@
+# Managed by Ansible (deploy/ansible).
+# Daily full pgBackRest base backup for the payments/game database. Runs pgBackRest
+# inside the postgres container (where the repository configuration and PGDATA live);
+# the continuous WAL between base backups is handled by the container's archive_command.
+# Installed only when pitr_enabled is set (see deploy/ansible/group_vars/all.yml).
+[Unit]
+Description=pgBackRest full base backup (scrabble)
+After=docker.service
+Requires=docker.service
+
+[Service]
+Type=oneshot
+# docker exec runs as the image's postgres user and inherits the container's PGBACKREST_*
+# environment (the S3 repository + cipher), so no repository config is needed on the host.
+ExecStart=/usr/bin/docker exec scrabble-postgres pgbackrest --stanza=scrabble --type=full backup

deploy/ansible/roles/main/templates/pgbackrest-backup.timer.j2

@@ -0,0 +1,12 @@
+# Managed by Ansible (deploy/ansible).
+# Fires the daily full base backup. Persistent=true catches up a run missed while the
+# host was down. Installed only when pitr_enabled is set.
+[Unit]
+Description=Daily pgBackRest full base backup (scrabble)
+
+[Timer]
+OnCalendar={{ pitr_backup_oncalendar }}
+Persistent=true
+
+[Install]
+WantedBy=timers.target

deploy/docker-compose.prod.yml

@@ -45,6 +45,43 @@ services:
           memory: 384M
 
   postgres:
+    image: ${REGISTRY:?set REGISTRY}/scrabble-postgres:${TAG:?set TAG}
+    # Continuous WAL archiving to S3 via pgBackRest (point-in-time recovery), on the prod
+    # main host only. archive_mode is gated by PGBACKREST_ARCHIVE_MODE (default "off"), so
+    # merging/redeploying this stack does NOT start archiving until the operator arms it:
+    # archiving stays inert — and cannot pile WAL onto the disk — until the S3 repository is
+    # set up, the stanza is created and the switch is flipped on. See deploy/README.md
+    # (point-in-time recovery — arming). Non-secret settings are inline; the endpoint,
+    # bucket, region, S3 keys and repository cipher passphrase come from the prod env.sh
+    # (deploy/write-prod-env.sh, PROD_ secrets/variables).
+    environment:
+      PGBACKREST_STANZA: scrabble
+      PGBACKREST_PG1_PATH: /var/lib/postgresql/data
+      PGBACKREST_REPO1_TYPE: s3
+      PGBACKREST_REPO1_PATH: /pgbackrest
+      PGBACKREST_REPO1_S3_URI_STYLE: path
+      PGBACKREST_REPO1_S3_ENDPOINT: ${PGBACKREST_S3_ENDPOINT:-}
+      PGBACKREST_REPO1_S3_PORT: ${PGBACKREST_S3_PORT:-443}
+      PGBACKREST_REPO1_S3_BUCKET: ${PGBACKREST_S3_BUCKET:-}
+      PGBACKREST_REPO1_S3_REGION: ${PGBACKREST_S3_REGION:-}
+      PGBACKREST_REPO1_S3_KEY: ${PGBACKREST_S3_KEY:-}
+      PGBACKREST_REPO1_S3_KEY_SECRET: ${PGBACKREST_S3_KEY_SECRET:-}
+      PGBACKREST_REPO1_CIPHER_TYPE: aes-256-cbc
+      PGBACKREST_REPO1_CIPHER_PASS: ${PGBACKREST_CIPHER_PASS:-}
+      PGBACKREST_REPO1_RETENTION_FULL: "30"
+      PGBACKREST_COMPRESS_TYPE: zst
+      PGBACKREST_LOG_LEVEL_CONSOLE: info
+      PGBACKREST_LOG_LEVEL_FILE: "off"
+    # archive_command is inert while archive_mode is off, so it is always present and only
+    # the mode switches. A forced 5-minute segment switch bounds the recovery point.
+    command:
+      - postgres
+      - -c
+      - archive_mode=${PGBACKREST_ARCHIVE_MODE:-off}
+      - -c
+      - "archive_command=pgbackrest --stanza=scrabble archive-push %p"
+      - -c
+      - archive_timeout=300
     deploy:
       resources:
         limits:

deploy/docker-compose.yml

@@ -38,7 +38,13 @@ x-logging: &default-logging
 services:
   postgres:
     container_name: scrabble-postgres
-    image: postgres:17-alpine
+    # Built from postgres:17-alpine + pgBackRest (deploy/postgres/Dockerfile) so the prod
+    # main host can archive WAL for point-in-time recovery. Archiving is switched on only by
+    # the prod overlay (docker-compose.prod.yml); on this contour the extra binary is present
+    # but idle and this is a plain postgres. See deploy/README.md (point-in-time recovery).
+    image: scrabble-postgres:latest
+    build:
+      context: postgres
     restart: unless-stopped
     logging: *default-logging
     environment:

deploy/grafana/provisioning/alerting/rules.yaml

@@ -212,3 +212,62 @@ groups:
                 - evaluator: { type: gt, params: [0.8] }
         labels: { severity: warning }
         annotations: { summary: 'Postgres using over 80% of max_connections.' }
+
+  # Continuous WAL archiving health (pgBackRest -> S3, point-in-time recovery). Both rules
+  # read the postgres_exporter's built-in pg_stat_archiver metrics and are absent/NaN-safe:
+  # on the test contour (and any host before archiving is armed) archive_mode is off, so
+  # failed_count stays 0 and last_archive_age is NaN — neither threshold trips.
+  - orgId: 1
+    name: scrabble-backup
+    folder: Alerts
+    interval: 1m
+    rules:
+      - uid: pg_archive_failing
+        title: WAL archiving failing
+        condition: C
+        for: 15m
+        noDataState: OK
+        execErrState: OK
+        data:
+          - refId: A
+            relativeTimeRange: { from: 900, to: 0 }
+            datasourceUid: prometheus
+            model:
+              refId: A
+              expr: increase(pg_stat_archiver_failed_count[15m])
+              instant: true
+          - refId: C
+            datasourceUid: __expr__
+            model:
+              refId: C
+              type: threshold
+              expression: A
+              conditions:
+                - evaluator: { type: gt, params: [0] }
+        labels: { severity: critical }
+        annotations: { summary: 'pgBackRest archive_command is failing — PITR is degrading and pg_wal can fill the disk. Run `docker exec scrabble-postgres pgbackrest --stanza=scrabble check`.' }
+
+      - uid: pg_archive_stalled
+        title: WAL archiving stalled
+        condition: C
+        for: 15m
+        noDataState: OK
+        execErrState: OK
+        data:
+          - refId: A
+            relativeTimeRange: { from: 300, to: 0 }
+            datasourceUid: prometheus
+            model:
+              refId: A
+              expr: pg_stat_archiver_last_archive_age
+              instant: true
+          - refId: C
+            datasourceUid: __expr__
+            model:
+              refId: C
+              type: threshold
+              expression: A
+              conditions:
+                - evaluator: { type: gt, params: [1800] }
+        labels: { severity: critical }
+        annotations: { summary: 'No WAL segment archived for over 30 minutes (archive_timeout is 5m) — archiving is stalled; the PITR recovery point is frozen and pg_wal may grow.' }

deploy/postgres/Dockerfile

@@ -0,0 +1,12 @@
+# Postgres for the Scrabble contour, extended with pgBackRest for continuous WAL
+# archiving and point-in-time recovery. pgBackRest must live in the database container
+# because Postgres runs archive_command in-process; the scheduled base backups and the
+# manual restore drills invoke the same binary through `docker exec`, inheriting the
+# repository configuration from the container environment.
+#
+# Archiving is enabled only on the production main host, where the prod overlay
+# (docker-compose.prod.yml) sets archive_mode and supplies the S3 repository. On the test
+# contour this image behaves exactly as a plain postgres:17-alpine. See deploy/README.md
+# (point-in-time recovery).
+FROM postgres:17-alpine
+RUN apk add --no-cache pgbackrest

deploy/prod-deploy.sh

@@ -140,7 +140,10 @@ if [ "$MIGRATION" = 1 ]; then
   echo "migration deploy: opening maintenance window (stopping the backend = the only writer)"
   dc stop backend
   dump="$DUMP_DIR/pre-$TAG-$(date +%Y%m%d-%H%M%S).sql"
-  if ! docker exec scrabble-postgres pg_dump -U "$PG_USER" -d "$PG_DB" -n backend > "$dump"; then
+  # Dump the WHOLE database (no -n): the payments schema — and any schema added later —
+  # must be in the belt-and-braces snapshot, not just backend. Restored manually into the
+  # same instance only if a migration turned out destructive (see deploy/README.md).
+  if ! docker exec scrabble-postgres pg_dump -U "$PG_USER" -d "$PG_DB" > "$dump"; then
     echo "pg_dump failed; restarting the old backend and aborting"
     dc start backend
     exit 1

deploy/write-prod-env.sh

@@ -73,4 +73,16 @@ export GF_SMTP_ENABLED='$GF_SMTP_ENABLED'
 export PUBLIC_BASE_URL='$PUBLIC_BASE_URL'
 export GATEWAY_HONEYTOKEN='$GATEWAY_HONEYTOKEN'
 export GATEWAY_ABUSE_BAN_ENABLED='true'
+# Continuous WAL archiving (pgBackRest -> S3) for point-in-time recovery. The artifact
+# ships disarmed: PGBACKREST_ARCHIVE_MODE defaults off, so archiving stays inert until the
+# operator sets the S3 repository values + secrets, creates the stanza and flips the switch
+# on (deploy/README.md). Empty repository values are harmless while the mode is off.
+export PGBACKREST_ARCHIVE_MODE='${PGBACKREST_ARCHIVE_MODE:-off}'
+export PGBACKREST_S3_ENDPOINT='$PGBACKREST_S3_ENDPOINT'
+export PGBACKREST_S3_PORT='${PGBACKREST_S3_PORT:-443}'
+export PGBACKREST_S3_BUCKET='$PGBACKREST_S3_BUCKET'
+export PGBACKREST_S3_REGION='$PGBACKREST_S3_REGION'
+export PGBACKREST_S3_KEY='$PGBACKREST_S3_KEY'
+export PGBACKREST_S3_KEY_SECRET='$PGBACKREST_S3_KEY_SECRET'
+export PGBACKREST_CIPHER_PASS='$PGBACKREST_CIPHER_PASS'
 EOF