diff --git a/DEPLOYMENT.md b/DEPLOYMENT.md
new file mode 100644
index 0000000..5ae6bd3
--- /dev/null
+++ b/DEPLOYMENT.md
@@ -0,0 +1,148 @@
+# Deploying RosterChief
+
+One server today, several later, with no code changes in between — only environment
+variables. This document is the runbook and, more usefully, the list of things that are
+specific to *this* app and will bite you if you treat it as a generic Django deploy.
+
+## The five things that make this deployment unusual
+
+**1. You need a wildcard TLS certificate, and that forces DNS-01.**
+Tenancy is subdomain-based (`ajax.rosterchief.app`), so the certificate must cover
+`*.rosterchief.app`. Let's Encrypt **will not issue a wildcard over HTTP-01** — only over
+DNS-01, which means the TLS terminator needs API access to your DNS zone. That is why
+`deploy/caddy/Dockerfile` builds Caddy *with* a DNS provider plugin, and why
+`CLOUDFLARE_API_TOKEN` is a required variable rather than a nicety. Swap the plugin
+(`caddy-dns/route53`, `caddy-dns/digitalocean`, …) if your DNS lives elsewhere.
+
+DNS needs two records, both pointing at the server:
+
+```
+A   rosterchief.app    -> <server ip>
+A   *.rosterchief.app  -> <server ip>
+```
+
+**2. Redis is not optional, even on one server.**
+`waffle` caches each feature flag's targeting in the Django cache, and `LocMemCache` is
+private to a single process. Under several gunicorn workers, toggling a feature in the
+control panel flushes **one** worker's cache while the others keep serving the stale flag —
+a feature that "sometimes doesn't turn on". A shared cache is the fix.
+
+**3. `SECURE_PROXY_SSL_HEADER` must be set, and Caddy must send the header.**
+Caddy terminates TLS, so without it Django believes every request is plain HTTP:
+`request.is_secure()` goes false, WebAuthn disagrees with the browser about the origin, and
+`SECURE_SSL_REDIRECT` becomes a redirect loop. Both halves are already wired (settings +
+`header_up X-Forwarded-Proto`); don't remove either.
+
+**4. Uploads must move to object storage before the second app server.**
+Club logos go to `MEDIA_ROOT` on local disk. On one box that is fine. On two, a logo
+uploaded to node A is a 404 on node B. Setting `AWS_STORAGE_BUCKET_NAME` switches the
+default storage to S3 — do it *before* you scale, not during.
+
+**5. PDF invoices need native libraries.**
+WeasyPrint binds to pango/cairo. The image installs them; a bare-metal deploy would need
+them too, and a Mac needs Homebrew. This is the main reason to run the container even in
+development if you touch invoicing.
+
+## First deploy
+
+```bash
+# 1. Configure
+cp .env.compose.example .env                 # read by docker compose
+cp .env.production.example .env.production   # read by Django
+python -c "import secrets; print(secrets.token_urlsafe(64))"   # -> DJANGO_SECRET_KEY
+
+# 2. Build and start
+docker compose build
+docker compose up -d db redis
+docker compose run --rm web python manage.py migrate
+docker compose run --rm web python manage.py createsuperuser
+docker compose up -d
+
+# 3. Verify
+curl -fsS https://rosterchief.app/healthz          # {"status": "ok", ...}
+docker compose run --rm web python manage.py check --deploy
+```
+
+`check --deploy` is what catches an env file that forgot the HTTPS flags: they default to
+**off** in code, because defaulting them to `not DEBUG` would redirect every test request to
+https and break the suite anywhere `DEBUG` is unset.
+
+The first `docker compose up` will take a minute or two: Caddy is provisioning the wildcard
+certificate over DNS-01, and DNS propagation is not instant. Watch it with
+`docker compose logs -f caddy`.
+
+## Migrations
+
+Deliberately **not** run by the container's entrypoint. With more than one web container they
+would race, and a starting gunicorn worker is a bad place to discover a failed migration.
+Run them once, explicitly, as part of the deploy:
+
+```bash
+docker compose build
+docker compose run --rm web python manage.py migrate
+docker compose up -d --no-deps web
+```
+
+## Scheduled jobs
+
+Two commands need to run on a schedule. Put them on the **host**, not in a container, and on
+**exactly one node** when you have several — three nodes archiving the same club is three
+emails to the same club.
+
+```cron
+# Bill: archive clubs unpaid past their grace period.
+# Run it WITHOUT --commit for the first week and read the output. The flag exists because
+# this switches off paying customers: a bad clock or a bad cron should cost you an email,
+# not a morning of angry clubs.
+0 6 * * *  cd /srv/rosterchief && docker compose run --rm web python manage.py archive_overdue_clubs --commit
+
+# Events: extend recurring series so the calendar never runs dry.
+0 3 * * *  cd /srv/rosterchief && docker compose run --rm web python manage.py extend_event_series
+```
+
+## Backups
+
+Two things carry state: Postgres and the uploads.
+
+```bash
+# Database
+docker compose exec -T db pg_dump -U rosterchief rosterchief | gzip > rosterchief-$(date +%F).sql.gz
+
+# Uploads — until they are on S3, in which case the bucket's own versioning is the backup.
+docker compose cp web:/app/media ./media-backup
+```
+
+Restore is `gunzip -c dump.sql.gz | docker compose exec -T db psql -U rosterchief rosterchief`.
+Test it once, now, rather than the first time you need it.
+
+## Going multi-server
+
+Nothing in the code changes. What changes is where the services live:
+
+| | one server | several |
+|---|---|---|
+| Postgres | `db` container | `DJANGO_DATABASE_URL` → your central Postgres |
+| Cache / flags | `redis` container | managed Redis (or your existing one) |
+| Uploads | local disk | **S3 bucket** (`AWS_STORAGE_BUCKET_NAME`) |
+| Static files | WhiteNoise, in the image | unchanged — that is why WhiteNoise is there |
+| Cron | host crontab | one node only |
+| TLS | Caddy on the box | load balancer, or Caddy on each node |
+
+Drop `db` and `redis` from `compose.yaml`, point the URLs at the central services, and run
+`web` on as many nodes as you like behind a load balancer pointed at `/healthz`.
+
+The health check tests the database *and* a cache round trip, not just that the process is
+listening — a node that cannot reach Postgres, or whose cache silently swallows writes, is
+not healthy, and a load balancer must not keep feeding it traffic.
+
+## Rollback
+
+Images are the unit of rollback. Tag on build, keep the last few, and:
+
+```bash
+docker compose up -d --no-deps web   # with the previous image tag
+```
+
+Migrations are the exception: they don't roll back with the image. Prefer additive migrations
+(add a column, deploy, backfill, then stop writing the old one) so that yesterday's image
+still runs against today's schema.
diff --git a/rosterchief/health.py b/rosterchief/health.py
new file mode 100644
index 0000000..aa12f1e
--- /dev/null
+++ b/rosterchief/health.py
@@ -0,0 +1,44 @@
+"""Liveness for the proxy today, for a load balancer later.
+
+Checks the two dependencies whose absence makes the app lie rather than fail: without the
+database it cannot serve anything, and without a shared cache the feature flags drift apart
+between workers. A health check that only proves the process is listening would call that
+healthy.
+"""
+
+from django.core.cache import cache
+from django.db import connection
+from django.http import JsonResponse
+from django.views.decorators.cache import never_cache
+
+PROBE_KEY = "healthz"
+
+
+@never_cache
+def healthz(request):
+    checks = {"database": _database(), "cache": _cache()}
+    healthy = all(checks.values())
+
+    return JsonResponse({"status": "ok" if healthy else "degraded", "checks": checks}, status=200 if healthy else 503)
+
+
+def _database() -> bool:
+    try:
+        with connection.cursor() as cursor:
+            cursor.execute("SELECT 1")
+            cursor.fetchone()
+    except Exception:
+        return False
+
+    return True
+
+
+def _cache() -> bool:
+    """A round trip, not a ping: a cache that accepts writes and returns nothing is worse
+    than one that is plainly down, because waffle would read every flag as unset."""
+    try:
+        cache.set(PROBE_KEY, "ok", 10)
+
+        return cache.get(PROBE_KEY) == "ok"
+    except Exception:
+        return False
diff --git a/rosterchief/tests.py b/rosterchief/tests.py
index 2dcf681..3a46149 100644
--- a/rosterchief/tests.py
+++ b/rosterchief/tests.py
@@ -1,7 +1,9 @@
 import importlib
+from unittest import mock
 
+from django.db.utils import OperationalError
 from django.test import SimpleTestCase, override_settings
-from django.urls import Resolver404, clear_url_caches, resolve
+from django.urls import Resolver404, clear_url_caches, resolve, reverse
 
 from . import urls
 
@@ -28,3 +30,51 @@ class BrowserReloadUrlTests(SimpleTestCase):
 
         with self.assertRaises(Resolver404):
             resolve("/__reload__/events/")
+
+
+class HealthCheckTests(SimpleTestCase):
+    databases = {"default"}
+
+    def test_it_reports_ok_when_the_database_and_cache_answer(self):
+        response = self.client.get(reverse("healthz"))
+
+        self.assertEqual(response.status_code, 200)
+        self.assertEqual(response.json(), {"status": "ok", "checks": {"database": True, "cache": True}})
+
+    def test_a_dead_database_is_a_503(self):
+        # 200 while the database is unreachable is worse than no health check at all: the load
+        # balancer would keep sending traffic to a node that cannot serve a single page.
+        with mock.patch("rosterchief.health.connection") as db:
+            db.cursor.side_effect = OperationalError("connection refused")
+
+            response = self.client.get(reverse("healthz"))
+
+        self.assertEqual(response.status_code, 503)
+        self.assertFalse(response.json()["checks"]["database"])
+
+    def test_a_cache_that_swallows_writes_is_unhealthy(self):
+        # Not a ping: a cache that accepts writes and returns nothing would have waffle read
+        # every feature flag as unset.
+        with mock.patch("rosterchief.health.cache") as broken:
+            broken.get.return_value = None
+
+            response = self.client.get(reverse("healthz"))
+
+        self.assertEqual(response.status_code, 503)
+        self.assertFalse(response.json()["checks"]["cache"])
+
+    def test_an_unreachable_cache_is_a_503(self):
+        # Distinct from the cache that answers wrongly above: here Redis is simply down, and
+        # the check must fail rather than raise its way to a 500.
+        with mock.patch("rosterchief.health.cache") as down:
+            down.set.side_effect = ConnectionError("redis is down")
+
+            response = self.client.get(reverse("healthz"))
+
+        self.assertEqual(response.status_code, 503)
+        self.assertFalse(response.json()["checks"]["cache"])
+
+    def test_it_is_never_cached(self):
+        response = self.client.get(reverse("healthz"))
+
+        self.assertIn("no-cache", response["Cache-Control"])
diff --git a/rosterchief/urls.py b/rosterchief/urls.py
index 581bff8..cd1c84d 100644
--- a/rosterchief/urls.py
+++ b/rosterchief/urls.py
@@ -15,7 +15,11 @@ from django.views.generic import RedirectView
 
 from club.views import root
 
+from .health import healthz
+
 urlpatterns = [
+    # No auth and no tenant: the proxy and the load balancer must reach it on any host.
+    path("healthz", healthz, name="healthz"),
     path("admin/login/", RedirectView.as_view(pattern_name="account_login", query_string=True), name="admin_login_redirect"),
     path("admin/", admin.site.urls),
     path("accounts/", include("allauth.urls")),