blumeops/containers/shower/default.nix

278 lines
9.8 KiB
Nix
Raw Normal View History

C1: deploy adelaide-baby-shower-app to ringtail k3s (#349) ## Summary Brings up the Adelaide / Heidi / Addie baby shower app on ringtail k3s with the public/private split that the app's hosting contract calls for: `shower.eblu.me` (public, via Fly proxy) and `shower.ops.eblu.me` (tailnet). App is consumed as a wheel from the Forgejo PyPI index — source lives at [`adelaide-baby-shower-app`](https://forge.eblu.me/eblume/adelaide-baby-shower-app). ### What's included - **ArgoCD app + manifests** under `argocd/manifests/shower/` (deployment, service, ProxyGroup ingress, ConfigMap for `DJANGO_DEBUG`/`DJANGO_ADMIN_URL`, ExternalSecret for `DJANGO_SECRET_KEY` from 1Password item `Shower (blumeops)`, NFS PV on sifaka, RWX media PVC, RWO local-path data PVC for SQLite). Recreate rollout because SQLite is single-writer. - **Public surface** (`fly/`): new `shower.eblu.me` server block proxying to `shower.ops.eblu.me`. `/admin/` returns 403 at the edge except `/admin/login/` and `/admin/logout/`, which are rate-limited via a new `shower_auth` zone. `X-Clacks-Overhead` on. GNU Terry Pratchett. - **fail2ban** filter (`shower-admin-login.conf`) matching 401/403/429 on `/admin/login/` and jail (`shower.conf`) with `maxretry=5/findtime=600/bantime=3600`. The `nginx-deny` action was generalized to take a per-jail `nginx_deny_file` so the shower has its own deny list (forge keeps using the legacy default). - **Caddy** route on indri (`shower.ops.eblu.me` → `https://shower.tail8d86e.ts.net`). - **Pulumi** Gandi CNAME `shower.eblu.me → blumeops-proxy.fly.dev.`. - **Grafana** APM dashboard `configmap-shower-apm.yaml` (request rate, error rate, failed admin login count, latency percentiles, bandwidth, access logs) mirroring `docs-apm.json` with a `host="shower.eblu.me"` filter. - **Container** `containers/shower/default.nix` — `dockerTools.buildLayeredImage` with a nixpkgs Python and a startup wrapper that creates `/app/data/.venv`, pip-installs `adelaide-baby-shower-app==1.0.0` from the forge PyPI index on first boot, runs migrations + collectstatic, and execs gunicorn. A `local_settings.py` shim pins `DATABASES.NAME`/`MEDIA_ROOT`/`STATIC_ROOT` to absolute paths so they don't end up in site-packages. - **Docs** runbook at `docs/how-to/operations/shower-app.md` linked from the apps registry, plus changelog fragments. ### Defense layers on the public surface 1. fly nginx geo+fail2ban `$shower_banned` (per-service deny list) 2. fly nginx `limit_req zone=shower_auth` (3 r/s per Fly-Client-IP) 3. django-axes (5 fails / 1h, keyed on username+ip_address) 4. edge `/admin/` block (returns 403 for anything that isn't login/logout) ## Prerequisites for the user to do (NOT in this PR) Halted on these per request — they touch shared/manual systems: - [x] **NFS share** on sifaka: `/volume1/shower`, NFS rule for ringtail RW, `chown 1000:1000` - [ ] **1Password item** `Shower (blumeops)` in the blumeops vault with a freshly minted `secret-key` field (`openssl rand -base64 48`) — do NOT reuse anything that has lived in git - [ ] **Container build**: `mise run container-build-and-release shower`, then update `images[].newTag` in `argocd/manifests/shower/kustomization.yaml` to the resulting `v1.0.0-<sha>-nix` - [x] **DNS**: `mise run dns-up` after merge - [x] **Fly cert**: `fly certs add shower.eblu.me -a blumeops-proxy` - [ ] **Caddy push**: `mise run provision-indri -- --tags caddy` - [ ] **Fly redeploy** to pick up the new nginx block + fail2ban jail: `mise run fly-deploy` - [ ] **ArgoCD sync**: `argocd app set shower --revision shower-app-deploy && argocd app sync shower` to test from this branch before merging ## Test plan - [ ] Container builds successfully on nix-container-builder runner - [ ] Pod starts, migrations run, gunicorn answers on :8000 - [ ] `kubectl --context=k3s-ringtail -n shower logs deploy/shower` clean - [ ] `curl -sf https://shower.ops.eblu.me/` returns the splash page (tailnet) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/` returns 200 (pre-DNS verification) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/admin/users/` returns 403 (edge block) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/admin/login/` returns a Django login response - [ ] After DNS is up: `curl -I https://shower.eblu.me/` returns 200 with `X-Clacks-Overhead` - [ ] Grafana dashboard "Shower APM" appears and starts showing traffic - [ ] `mise run services-check` passes Reviewed-on: https://forge.eblu.me/eblume/blumeops/pulls/349
2026-05-11 13:47:18 -07:00
# Nix-built shower app container — Adelaide / Heidi / Addie baby shower.
#
# The app is published as a wheel to the Forgejo PyPI index at
C1: deploy shower v1.1.0 (phases + guest memories) (#354) ## Summary Deploys `adelaide-baby-shower-app` **v1.1.0** to ringtail k3s. ### App changes (since v1.0.2) - **Four-phase `ShowerState`** replaces the boolean `locked` flag — `pre_event` → `party` → `prizes_locked` → `event_locked` — with a backfill migration that maps `locked=True → pre_event`, `locked=False → party`. - **Guest memories**: append-only photos + comments panel where guests can leave notes for the baby. Adds `GuestPhoto` + `GuestComment` models with file-extension validators and a max-size validator; new `shower.imaging` module for thumbnail generation. - **Admin + QR polish**: configurable host link, fixed "View Site" URL, guest-facing QR copy improvements, contest tweaks. Three Django migrations run automatically in the entrypoint against the SQLite PV: - `0009_shower_phase` - `0010_guest_memories` - `0011_book_description` No ConfigMap / env-var changes. The deploy uses `strategy: Recreate` with a single replica, so the old pod releases the data PVC before the new one mounts it and runs migrations. ### Container build changes The v1.1.0 tag exposed a latent issue with the Forgejo PyPI install path: - The recent commit [2d38418e](https://forge.eblu.me/eblume/blumeops/commit/2d38418e) closed the forge package leak at the Fly edge by blocking `/api/packages/*` publicly. - Forgejo's PyPI simple index returns absolute file URLs hardcoded to its public `ROOT_URL` (`forge.eblu.me`), so pip-installing from the tailnet index URL still tries to download from `forge.eblu.me` → 403. - Previous shower builds escaped this because their FOD outputs were already in the nix store; bumping to a new version forced a fresh pip run that hit the block. Fix mirrors what we already do for the sdist: both wheel and sdist are pulled via direct `fetchurl` against `forge.ops.eblu.me`, then the wheel is copied to TMPDIR under its clean filename (nix store path's hash prefix breaks pip's wheel-filename parser) and handed to pip as a local path. The forge `--extra-index-url` is no longer needed. FOD outputHash pinned to `sha256-kTNOswobtkgyQmmqbQM8XO4vvaGg57nCuuZGbNXb0NM=` from run 547. Image: `registry.ops.eblu.me/blumeops/shower:v1.1.0-444ff91-nix`. ### Adjacent finding (already handled) The ringtail `gitea-runner-nix_container_builder` systemd unit was left `inactive` after the recent `provision-ringtail` (matches the known `sshd-restart-hangs-mux` lesson — the rebuild changed the unit's PATH closure + config.yaml, systemd stopped it, then the playbook hung before the activation could restart it). Manually started; the existing memory `lesson_provision_ringtail_ssh_hang.md` was extended to mention the runner as the canary service to check after provisions. ## Test plan - [ ] `argocd app diff shower --revision shower-v1.1.0` — review the manifest change - [ ] `argocd app set shower --revision shower-v1.1.0 && argocd app sync shower` - [ ] `kubectl --context=k3s-ringtail logs -n shower deploy/shower` — confirm migrations 0009/0010/0011 applied, no errors - [ ] Hit `https://shower.ops.eblu.me/` (tailnet) — splash page renders, phase indicator visible - [ ] Hit `https://shower.ops.eblu.me/host/` — host console loads, phase dropdown shows the four states - [ ] Hit `https://shower.eblu.me/` (public via Fly) — splash page still served - [ ] After merge: `argocd app set shower --revision main && argocd app sync shower` Reviewed-on: https://forge.eblu.me/eblume/blumeops/pulls/354
2026-05-11 20:08:03 -07:00
# https://forge.ops.eblu.me/api/packages/eblume/pypi/ (tailnet-only — the
# public forge.eblu.me /api/packages/* surface is blocked at the Fly edge).
# We can't point pip at Forgejo's simple index even from the tailnet,
# because Forgejo's index returns absolute file URLs hardcoded to its
# public ROOT_URL (forge.eblu.me), which then 403s. So both the wheel and
# the sdist are pulled by direct `fetchurl` against forge.ops.eblu.me, and
# the wheel is then handed to `pip install` as a local path; transitive
# deps come from pypi.ops.eblu.me. Build runs on the nix-container-builder
# runner (ringtail, amd64) so the image is native.
C1: deploy adelaide-baby-shower-app to ringtail k3s (#349) ## Summary Brings up the Adelaide / Heidi / Addie baby shower app on ringtail k3s with the public/private split that the app's hosting contract calls for: `shower.eblu.me` (public, via Fly proxy) and `shower.ops.eblu.me` (tailnet). App is consumed as a wheel from the Forgejo PyPI index — source lives at [`adelaide-baby-shower-app`](https://forge.eblu.me/eblume/adelaide-baby-shower-app). ### What's included - **ArgoCD app + manifests** under `argocd/manifests/shower/` (deployment, service, ProxyGroup ingress, ConfigMap for `DJANGO_DEBUG`/`DJANGO_ADMIN_URL`, ExternalSecret for `DJANGO_SECRET_KEY` from 1Password item `Shower (blumeops)`, NFS PV on sifaka, RWX media PVC, RWO local-path data PVC for SQLite). Recreate rollout because SQLite is single-writer. - **Public surface** (`fly/`): new `shower.eblu.me` server block proxying to `shower.ops.eblu.me`. `/admin/` returns 403 at the edge except `/admin/login/` and `/admin/logout/`, which are rate-limited via a new `shower_auth` zone. `X-Clacks-Overhead` on. GNU Terry Pratchett. - **fail2ban** filter (`shower-admin-login.conf`) matching 401/403/429 on `/admin/login/` and jail (`shower.conf`) with `maxretry=5/findtime=600/bantime=3600`. The `nginx-deny` action was generalized to take a per-jail `nginx_deny_file` so the shower has its own deny list (forge keeps using the legacy default). - **Caddy** route on indri (`shower.ops.eblu.me` → `https://shower.tail8d86e.ts.net`). - **Pulumi** Gandi CNAME `shower.eblu.me → blumeops-proxy.fly.dev.`. - **Grafana** APM dashboard `configmap-shower-apm.yaml` (request rate, error rate, failed admin login count, latency percentiles, bandwidth, access logs) mirroring `docs-apm.json` with a `host="shower.eblu.me"` filter. - **Container** `containers/shower/default.nix` — `dockerTools.buildLayeredImage` with a nixpkgs Python and a startup wrapper that creates `/app/data/.venv`, pip-installs `adelaide-baby-shower-app==1.0.0` from the forge PyPI index on first boot, runs migrations + collectstatic, and execs gunicorn. A `local_settings.py` shim pins `DATABASES.NAME`/`MEDIA_ROOT`/`STATIC_ROOT` to absolute paths so they don't end up in site-packages. - **Docs** runbook at `docs/how-to/operations/shower-app.md` linked from the apps registry, plus changelog fragments. ### Defense layers on the public surface 1. fly nginx geo+fail2ban `$shower_banned` (per-service deny list) 2. fly nginx `limit_req zone=shower_auth` (3 r/s per Fly-Client-IP) 3. django-axes (5 fails / 1h, keyed on username+ip_address) 4. edge `/admin/` block (returns 403 for anything that isn't login/logout) ## Prerequisites for the user to do (NOT in this PR) Halted on these per request — they touch shared/manual systems: - [x] **NFS share** on sifaka: `/volume1/shower`, NFS rule for ringtail RW, `chown 1000:1000` - [ ] **1Password item** `Shower (blumeops)` in the blumeops vault with a freshly minted `secret-key` field (`openssl rand -base64 48`) — do NOT reuse anything that has lived in git - [ ] **Container build**: `mise run container-build-and-release shower`, then update `images[].newTag` in `argocd/manifests/shower/kustomization.yaml` to the resulting `v1.0.0-<sha>-nix` - [x] **DNS**: `mise run dns-up` after merge - [x] **Fly cert**: `fly certs add shower.eblu.me -a blumeops-proxy` - [ ] **Caddy push**: `mise run provision-indri -- --tags caddy` - [ ] **Fly redeploy** to pick up the new nginx block + fail2ban jail: `mise run fly-deploy` - [ ] **ArgoCD sync**: `argocd app set shower --revision shower-app-deploy && argocd app sync shower` to test from this branch before merging ## Test plan - [ ] Container builds successfully on nix-container-builder runner - [ ] Pod starts, migrations run, gunicorn answers on :8000 - [ ] `kubectl --context=k3s-ringtail -n shower logs deploy/shower` clean - [ ] `curl -sf https://shower.ops.eblu.me/` returns the splash page (tailnet) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/` returns 200 (pre-DNS verification) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/admin/users/` returns 403 (edge block) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/admin/login/` returns a Django login response - [ ] After DNS is up: `curl -I https://shower.eblu.me/` returns 200 with `X-Clacks-Overhead` - [ ] Grafana dashboard "Shower APM" appears and starts showing traffic - [ ] `mise run services-check` passes Reviewed-on: https://forge.eblu.me/eblume/blumeops/pulls/349
2026-05-11 13:47:18 -07:00
#
# Going through pip-install-target rather than nixpkgs Python packages
# sidesteps two issues we hit going through `python.pkgs.buildPythonPackage`:
# 1. python314Packages.django still aliases to Django 4.2 LTS, which
# doesn't support Python 3.14 at all.
# 2. django-axes pulls selenium + browser fonts into its check phase
# and the nix sandbox can't provide those.
#
# To bump the version:
# 1. Update `version` below.
# 2. Set `outputHash` to `pkgs.lib.fakeHash`, run the build, copy the
# real hash out of the error, and commit it.
{ pkgs ? import <nixpkgs> { } }:
let
version = "1.1.3";
C1: deploy adelaide-baby-shower-app to ringtail k3s (#349) ## Summary Brings up the Adelaide / Heidi / Addie baby shower app on ringtail k3s with the public/private split that the app's hosting contract calls for: `shower.eblu.me` (public, via Fly proxy) and `shower.ops.eblu.me` (tailnet). App is consumed as a wheel from the Forgejo PyPI index — source lives at [`adelaide-baby-shower-app`](https://forge.eblu.me/eblume/adelaide-baby-shower-app). ### What's included - **ArgoCD app + manifests** under `argocd/manifests/shower/` (deployment, service, ProxyGroup ingress, ConfigMap for `DJANGO_DEBUG`/`DJANGO_ADMIN_URL`, ExternalSecret for `DJANGO_SECRET_KEY` from 1Password item `Shower (blumeops)`, NFS PV on sifaka, RWX media PVC, RWO local-path data PVC for SQLite). Recreate rollout because SQLite is single-writer. - **Public surface** (`fly/`): new `shower.eblu.me` server block proxying to `shower.ops.eblu.me`. `/admin/` returns 403 at the edge except `/admin/login/` and `/admin/logout/`, which are rate-limited via a new `shower_auth` zone. `X-Clacks-Overhead` on. GNU Terry Pratchett. - **fail2ban** filter (`shower-admin-login.conf`) matching 401/403/429 on `/admin/login/` and jail (`shower.conf`) with `maxretry=5/findtime=600/bantime=3600`. The `nginx-deny` action was generalized to take a per-jail `nginx_deny_file` so the shower has its own deny list (forge keeps using the legacy default). - **Caddy** route on indri (`shower.ops.eblu.me` → `https://shower.tail8d86e.ts.net`). - **Pulumi** Gandi CNAME `shower.eblu.me → blumeops-proxy.fly.dev.`. - **Grafana** APM dashboard `configmap-shower-apm.yaml` (request rate, error rate, failed admin login count, latency percentiles, bandwidth, access logs) mirroring `docs-apm.json` with a `host="shower.eblu.me"` filter. - **Container** `containers/shower/default.nix` — `dockerTools.buildLayeredImage` with a nixpkgs Python and a startup wrapper that creates `/app/data/.venv`, pip-installs `adelaide-baby-shower-app==1.0.0` from the forge PyPI index on first boot, runs migrations + collectstatic, and execs gunicorn. A `local_settings.py` shim pins `DATABASES.NAME`/`MEDIA_ROOT`/`STATIC_ROOT` to absolute paths so they don't end up in site-packages. - **Docs** runbook at `docs/how-to/operations/shower-app.md` linked from the apps registry, plus changelog fragments. ### Defense layers on the public surface 1. fly nginx geo+fail2ban `$shower_banned` (per-service deny list) 2. fly nginx `limit_req zone=shower_auth` (3 r/s per Fly-Client-IP) 3. django-axes (5 fails / 1h, keyed on username+ip_address) 4. edge `/admin/` block (returns 403 for anything that isn't login/logout) ## Prerequisites for the user to do (NOT in this PR) Halted on these per request — they touch shared/manual systems: - [x] **NFS share** on sifaka: `/volume1/shower`, NFS rule for ringtail RW, `chown 1000:1000` - [ ] **1Password item** `Shower (blumeops)` in the blumeops vault with a freshly minted `secret-key` field (`openssl rand -base64 48`) — do NOT reuse anything that has lived in git - [ ] **Container build**: `mise run container-build-and-release shower`, then update `images[].newTag` in `argocd/manifests/shower/kustomization.yaml` to the resulting `v1.0.0-<sha>-nix` - [x] **DNS**: `mise run dns-up` after merge - [x] **Fly cert**: `fly certs add shower.eblu.me -a blumeops-proxy` - [ ] **Caddy push**: `mise run provision-indri -- --tags caddy` - [ ] **Fly redeploy** to pick up the new nginx block + fail2ban jail: `mise run fly-deploy` - [ ] **ArgoCD sync**: `argocd app set shower --revision shower-app-deploy && argocd app sync shower` to test from this branch before merging ## Test plan - [ ] Container builds successfully on nix-container-builder runner - [ ] Pod starts, migrations run, gunicorn answers on :8000 - [ ] `kubectl --context=k3s-ringtail -n shower logs deploy/shower` clean - [ ] `curl -sf https://shower.ops.eblu.me/` returns the splash page (tailnet) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/` returns 200 (pre-DNS verification) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/admin/users/` returns 403 (edge block) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/admin/login/` returns a Django login response - [ ] After DNS is up: `curl -I https://shower.eblu.me/` returns 200 with `X-Clacks-Overhead` - [ ] Grafana dashboard "Shower APM" appears and starts showing traffic - [ ] `mise run services-check` passes Reviewed-on: https://forge.eblu.me/eblume/blumeops/pulls/349
2026-05-11 13:47:18 -07:00
python = pkgs.python314;
# The repo's top-level static/ directory (vendored Sortable + cropper
# JS/CSS, prize placeholder SVG) isn't shipped in the wheel — hatchling
# only packages config/ and shower/, leaving the repo-root static/
# behind. Pull the sdist (which contains the full source tree) and
# extract just the static/ subtree into the image as /app/static.
# local_settings adds it to STATICFILES_DIRS so collectstatic at boot
# picks it up alongside the Django admin's static files.
#
# Fetched from forge.ops.eblu.me (tailnet) because /api/packages/* is
# blocked at the fly edge — see fly/nginx.conf forge.eblu.me block.
# Hash is the upstream sha256 from forge PyPI's simple index.
showerSdist = pkgs.fetchurl {
name = "adelaide_baby_shower_app-${version}.tar.gz";
url = "https://forge.ops.eblu.me/api/packages/eblume/pypi/files/adelaide-baby-shower-app/${version}/adelaide_baby_shower_app-${version}.tar.gz";
hash = "sha256-a3rCwEdOB+rnYXqsWDifyltpyKUgkOj0ikWB+WGQYKE=";
C1: deploy shower v1.1.0 (phases + guest memories) (#354) ## Summary Deploys `adelaide-baby-shower-app` **v1.1.0** to ringtail k3s. ### App changes (since v1.0.2) - **Four-phase `ShowerState`** replaces the boolean `locked` flag — `pre_event` → `party` → `prizes_locked` → `event_locked` — with a backfill migration that maps `locked=True → pre_event`, `locked=False → party`. - **Guest memories**: append-only photos + comments panel where guests can leave notes for the baby. Adds `GuestPhoto` + `GuestComment` models with file-extension validators and a max-size validator; new `shower.imaging` module for thumbnail generation. - **Admin + QR polish**: configurable host link, fixed "View Site" URL, guest-facing QR copy improvements, contest tweaks. Three Django migrations run automatically in the entrypoint against the SQLite PV: - `0009_shower_phase` - `0010_guest_memories` - `0011_book_description` No ConfigMap / env-var changes. The deploy uses `strategy: Recreate` with a single replica, so the old pod releases the data PVC before the new one mounts it and runs migrations. ### Container build changes The v1.1.0 tag exposed a latent issue with the Forgejo PyPI install path: - The recent commit [2d38418e](https://forge.eblu.me/eblume/blumeops/commit/2d38418e) closed the forge package leak at the Fly edge by blocking `/api/packages/*` publicly. - Forgejo's PyPI simple index returns absolute file URLs hardcoded to its public `ROOT_URL` (`forge.eblu.me`), so pip-installing from the tailnet index URL still tries to download from `forge.eblu.me` → 403. - Previous shower builds escaped this because their FOD outputs were already in the nix store; bumping to a new version forced a fresh pip run that hit the block. Fix mirrors what we already do for the sdist: both wheel and sdist are pulled via direct `fetchurl` against `forge.ops.eblu.me`, then the wheel is copied to TMPDIR under its clean filename (nix store path's hash prefix breaks pip's wheel-filename parser) and handed to pip as a local path. The forge `--extra-index-url` is no longer needed. FOD outputHash pinned to `sha256-kTNOswobtkgyQmmqbQM8XO4vvaGg57nCuuZGbNXb0NM=` from run 547. Image: `registry.ops.eblu.me/blumeops/shower:v1.1.0-444ff91-nix`. ### Adjacent finding (already handled) The ringtail `gitea-runner-nix_container_builder` systemd unit was left `inactive` after the recent `provision-ringtail` (matches the known `sshd-restart-hangs-mux` lesson — the rebuild changed the unit's PATH closure + config.yaml, systemd stopped it, then the playbook hung before the activation could restart it). Manually started; the existing memory `lesson_provision_ringtail_ssh_hang.md` was extended to mention the runner as the canary service to check after provisions. ## Test plan - [ ] `argocd app diff shower --revision shower-v1.1.0` — review the manifest change - [ ] `argocd app set shower --revision shower-v1.1.0 && argocd app sync shower` - [ ] `kubectl --context=k3s-ringtail logs -n shower deploy/shower` — confirm migrations 0009/0010/0011 applied, no errors - [ ] Hit `https://shower.ops.eblu.me/` (tailnet) — splash page renders, phase indicator visible - [ ] Hit `https://shower.ops.eblu.me/host/` — host console loads, phase dropdown shows the four states - [ ] Hit `https://shower.eblu.me/` (public via Fly) — splash page still served - [ ] After merge: `argocd app set shower --revision main && argocd app sync shower` Reviewed-on: https://forge.eblu.me/eblume/blumeops/pulls/354
2026-05-11 20:08:03 -07:00
};
# Wheel pulled from forge.ops.eblu.me (tailnet) for the same reason the
# sdist is: Forgejo's PyPI simple index would return forge.eblu.me URLs
# that the Fly edge 403s on /api/packages/*. We hand this path to pip
# below so it never touches the forge index at all.
showerWheel = pkgs.fetchurl {
name = "adelaide_baby_shower_app-${version}-py3-none-any.whl";
url = "https://forge.ops.eblu.me/api/packages/eblume/pypi/files/adelaide-baby-shower-app/${version}/adelaide_baby_shower_app-${version}-py3-none-any.whl";
hash = "sha256-a6j91gBigG4IzE2DVTBntnZ46Yrx9b5PgHn+Uro98Tk=";
C1: deploy adelaide-baby-shower-app to ringtail k3s (#349) ## Summary Brings up the Adelaide / Heidi / Addie baby shower app on ringtail k3s with the public/private split that the app's hosting contract calls for: `shower.eblu.me` (public, via Fly proxy) and `shower.ops.eblu.me` (tailnet). App is consumed as a wheel from the Forgejo PyPI index — source lives at [`adelaide-baby-shower-app`](https://forge.eblu.me/eblume/adelaide-baby-shower-app). ### What's included - **ArgoCD app + manifests** under `argocd/manifests/shower/` (deployment, service, ProxyGroup ingress, ConfigMap for `DJANGO_DEBUG`/`DJANGO_ADMIN_URL`, ExternalSecret for `DJANGO_SECRET_KEY` from 1Password item `Shower (blumeops)`, NFS PV on sifaka, RWX media PVC, RWO local-path data PVC for SQLite). Recreate rollout because SQLite is single-writer. - **Public surface** (`fly/`): new `shower.eblu.me` server block proxying to `shower.ops.eblu.me`. `/admin/` returns 403 at the edge except `/admin/login/` and `/admin/logout/`, which are rate-limited via a new `shower_auth` zone. `X-Clacks-Overhead` on. GNU Terry Pratchett. - **fail2ban** filter (`shower-admin-login.conf`) matching 401/403/429 on `/admin/login/` and jail (`shower.conf`) with `maxretry=5/findtime=600/bantime=3600`. The `nginx-deny` action was generalized to take a per-jail `nginx_deny_file` so the shower has its own deny list (forge keeps using the legacy default). - **Caddy** route on indri (`shower.ops.eblu.me` → `https://shower.tail8d86e.ts.net`). - **Pulumi** Gandi CNAME `shower.eblu.me → blumeops-proxy.fly.dev.`. - **Grafana** APM dashboard `configmap-shower-apm.yaml` (request rate, error rate, failed admin login count, latency percentiles, bandwidth, access logs) mirroring `docs-apm.json` with a `host="shower.eblu.me"` filter. - **Container** `containers/shower/default.nix` — `dockerTools.buildLayeredImage` with a nixpkgs Python and a startup wrapper that creates `/app/data/.venv`, pip-installs `adelaide-baby-shower-app==1.0.0` from the forge PyPI index on first boot, runs migrations + collectstatic, and execs gunicorn. A `local_settings.py` shim pins `DATABASES.NAME`/`MEDIA_ROOT`/`STATIC_ROOT` to absolute paths so they don't end up in site-packages. - **Docs** runbook at `docs/how-to/operations/shower-app.md` linked from the apps registry, plus changelog fragments. ### Defense layers on the public surface 1. fly nginx geo+fail2ban `$shower_banned` (per-service deny list) 2. fly nginx `limit_req zone=shower_auth` (3 r/s per Fly-Client-IP) 3. django-axes (5 fails / 1h, keyed on username+ip_address) 4. edge `/admin/` block (returns 403 for anything that isn't login/logout) ## Prerequisites for the user to do (NOT in this PR) Halted on these per request — they touch shared/manual systems: - [x] **NFS share** on sifaka: `/volume1/shower`, NFS rule for ringtail RW, `chown 1000:1000` - [ ] **1Password item** `Shower (blumeops)` in the blumeops vault with a freshly minted `secret-key` field (`openssl rand -base64 48`) — do NOT reuse anything that has lived in git - [ ] **Container build**: `mise run container-build-and-release shower`, then update `images[].newTag` in `argocd/manifests/shower/kustomization.yaml` to the resulting `v1.0.0-<sha>-nix` - [x] **DNS**: `mise run dns-up` after merge - [x] **Fly cert**: `fly certs add shower.eblu.me -a blumeops-proxy` - [ ] **Caddy push**: `mise run provision-indri -- --tags caddy` - [ ] **Fly redeploy** to pick up the new nginx block + fail2ban jail: `mise run fly-deploy` - [ ] **ArgoCD sync**: `argocd app set shower --revision shower-app-deploy && argocd app sync shower` to test from this branch before merging ## Test plan - [ ] Container builds successfully on nix-container-builder runner - [ ] Pod starts, migrations run, gunicorn answers on :8000 - [ ] `kubectl --context=k3s-ringtail -n shower logs deploy/shower` clean - [ ] `curl -sf https://shower.ops.eblu.me/` returns the splash page (tailnet) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/` returns 200 (pre-DNS verification) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/admin/users/` returns 403 (edge block) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/admin/login/` returns a Django login response - [ ] After DNS is up: `curl -I https://shower.eblu.me/` returns 200 with `X-Clacks-Overhead` - [ ] Grafana dashboard "Shower APM" appears and starts showing traffic - [ ] `mise run services-check` passes Reviewed-on: https://forge.eblu.me/eblume/blumeops/pulls/349
2026-05-11 13:47:18 -07:00
};
staticAssets = pkgs.runCommand "shower-static-assets-${version}" { } ''
${pkgs.gnutar}/bin/tar -xzf ${showerSdist} -C $TMPDIR
cp -r $TMPDIR/adelaide_baby_shower_app-${version}/static $out
'';
# Fixed-output derivation: pip-installs the app wheel + every transitive
# dep into a single target dir. FODs get network access in exchange for
# a pinned output hash, which means the whole dependency closure is
# immutable across rebuilds.
pyDepsFOD = pkgs.stdenv.mkDerivation {
pname = "shower-python-deps-fod";
inherit version;
dontUnpack = true;
nativeBuildInputs = [ python pkgs.cacert pkgs.removeReferencesTo ];
buildPhase = ''
runHook preBuild
export HOME=$TMPDIR
export SSL_CERT_FILE=${pkgs.cacert}/etc/ssl/certs/ca-bundle.crt
export PIP_DISABLE_PIP_VERSION_CHECK=1
${python}/bin/python -m venv "$TMPDIR/venv"
"$TMPDIR/venv/bin/pip" install --upgrade pip
C1: deploy shower v1.1.0 (phases + guest memories) (#354) ## Summary Deploys `adelaide-baby-shower-app` **v1.1.0** to ringtail k3s. ### App changes (since v1.0.2) - **Four-phase `ShowerState`** replaces the boolean `locked` flag — `pre_event` → `party` → `prizes_locked` → `event_locked` — with a backfill migration that maps `locked=True → pre_event`, `locked=False → party`. - **Guest memories**: append-only photos + comments panel where guests can leave notes for the baby. Adds `GuestPhoto` + `GuestComment` models with file-extension validators and a max-size validator; new `shower.imaging` module for thumbnail generation. - **Admin + QR polish**: configurable host link, fixed "View Site" URL, guest-facing QR copy improvements, contest tweaks. Three Django migrations run automatically in the entrypoint against the SQLite PV: - `0009_shower_phase` - `0010_guest_memories` - `0011_book_description` No ConfigMap / env-var changes. The deploy uses `strategy: Recreate` with a single replica, so the old pod releases the data PVC before the new one mounts it and runs migrations. ### Container build changes The v1.1.0 tag exposed a latent issue with the Forgejo PyPI install path: - The recent commit [2d38418e](https://forge.eblu.me/eblume/blumeops/commit/2d38418e) closed the forge package leak at the Fly edge by blocking `/api/packages/*` publicly. - Forgejo's PyPI simple index returns absolute file URLs hardcoded to its public `ROOT_URL` (`forge.eblu.me`), so pip-installing from the tailnet index URL still tries to download from `forge.eblu.me` → 403. - Previous shower builds escaped this because their FOD outputs were already in the nix store; bumping to a new version forced a fresh pip run that hit the block. Fix mirrors what we already do for the sdist: both wheel and sdist are pulled via direct `fetchurl` against `forge.ops.eblu.me`, then the wheel is copied to TMPDIR under its clean filename (nix store path's hash prefix breaks pip's wheel-filename parser) and handed to pip as a local path. The forge `--extra-index-url` is no longer needed. FOD outputHash pinned to `sha256-kTNOswobtkgyQmmqbQM8XO4vvaGg57nCuuZGbNXb0NM=` from run 547. Image: `registry.ops.eblu.me/blumeops/shower:v1.1.0-444ff91-nix`. ### Adjacent finding (already handled) The ringtail `gitea-runner-nix_container_builder` systemd unit was left `inactive` after the recent `provision-ringtail` (matches the known `sshd-restart-hangs-mux` lesson — the rebuild changed the unit's PATH closure + config.yaml, systemd stopped it, then the playbook hung before the activation could restart it). Manually started; the existing memory `lesson_provision_ringtail_ssh_hang.md` was extended to mention the runner as the canary service to check after provisions. ## Test plan - [ ] `argocd app diff shower --revision shower-v1.1.0` — review the manifest change - [ ] `argocd app set shower --revision shower-v1.1.0 && argocd app sync shower` - [ ] `kubectl --context=k3s-ringtail logs -n shower deploy/shower` — confirm migrations 0009/0010/0011 applied, no errors - [ ] Hit `https://shower.ops.eblu.me/` (tailnet) — splash page renders, phase indicator visible - [ ] Hit `https://shower.ops.eblu.me/host/` — host console loads, phase dropdown shows the four states - [ ] Hit `https://shower.eblu.me/` (public via Fly) — splash page still served - [ ] After merge: `argocd app set shower --revision main && argocd app sync shower` Reviewed-on: https://forge.eblu.me/eblume/blumeops/pulls/354
2026-05-11 20:08:03 -07:00
# Nix store paths embed a 32-char hash prefix, which pip's wheel
# filename parser rejects ("Invalid wheel filename"). Copy to a
# clean filename in TMPDIR before installing.
cp ${showerWheel} "$TMPDIR/${showerWheel.name}"
C1: deploy adelaide-baby-shower-app to ringtail k3s (#349) ## Summary Brings up the Adelaide / Heidi / Addie baby shower app on ringtail k3s with the public/private split that the app's hosting contract calls for: `shower.eblu.me` (public, via Fly proxy) and `shower.ops.eblu.me` (tailnet). App is consumed as a wheel from the Forgejo PyPI index — source lives at [`adelaide-baby-shower-app`](https://forge.eblu.me/eblume/adelaide-baby-shower-app). ### What's included - **ArgoCD app + manifests** under `argocd/manifests/shower/` (deployment, service, ProxyGroup ingress, ConfigMap for `DJANGO_DEBUG`/`DJANGO_ADMIN_URL`, ExternalSecret for `DJANGO_SECRET_KEY` from 1Password item `Shower (blumeops)`, NFS PV on sifaka, RWX media PVC, RWO local-path data PVC for SQLite). Recreate rollout because SQLite is single-writer. - **Public surface** (`fly/`): new `shower.eblu.me` server block proxying to `shower.ops.eblu.me`. `/admin/` returns 403 at the edge except `/admin/login/` and `/admin/logout/`, which are rate-limited via a new `shower_auth` zone. `X-Clacks-Overhead` on. GNU Terry Pratchett. - **fail2ban** filter (`shower-admin-login.conf`) matching 401/403/429 on `/admin/login/` and jail (`shower.conf`) with `maxretry=5/findtime=600/bantime=3600`. The `nginx-deny` action was generalized to take a per-jail `nginx_deny_file` so the shower has its own deny list (forge keeps using the legacy default). - **Caddy** route on indri (`shower.ops.eblu.me` → `https://shower.tail8d86e.ts.net`). - **Pulumi** Gandi CNAME `shower.eblu.me → blumeops-proxy.fly.dev.`. - **Grafana** APM dashboard `configmap-shower-apm.yaml` (request rate, error rate, failed admin login count, latency percentiles, bandwidth, access logs) mirroring `docs-apm.json` with a `host="shower.eblu.me"` filter. - **Container** `containers/shower/default.nix` — `dockerTools.buildLayeredImage` with a nixpkgs Python and a startup wrapper that creates `/app/data/.venv`, pip-installs `adelaide-baby-shower-app==1.0.0` from the forge PyPI index on first boot, runs migrations + collectstatic, and execs gunicorn. A `local_settings.py` shim pins `DATABASES.NAME`/`MEDIA_ROOT`/`STATIC_ROOT` to absolute paths so they don't end up in site-packages. - **Docs** runbook at `docs/how-to/operations/shower-app.md` linked from the apps registry, plus changelog fragments. ### Defense layers on the public surface 1. fly nginx geo+fail2ban `$shower_banned` (per-service deny list) 2. fly nginx `limit_req zone=shower_auth` (3 r/s per Fly-Client-IP) 3. django-axes (5 fails / 1h, keyed on username+ip_address) 4. edge `/admin/` block (returns 403 for anything that isn't login/logout) ## Prerequisites for the user to do (NOT in this PR) Halted on these per request — they touch shared/manual systems: - [x] **NFS share** on sifaka: `/volume1/shower`, NFS rule for ringtail RW, `chown 1000:1000` - [ ] **1Password item** `Shower (blumeops)` in the blumeops vault with a freshly minted `secret-key` field (`openssl rand -base64 48`) — do NOT reuse anything that has lived in git - [ ] **Container build**: `mise run container-build-and-release shower`, then update `images[].newTag` in `argocd/manifests/shower/kustomization.yaml` to the resulting `v1.0.0-<sha>-nix` - [x] **DNS**: `mise run dns-up` after merge - [x] **Fly cert**: `fly certs add shower.eblu.me -a blumeops-proxy` - [ ] **Caddy push**: `mise run provision-indri -- --tags caddy` - [ ] **Fly redeploy** to pick up the new nginx block + fail2ban jail: `mise run fly-deploy` - [ ] **ArgoCD sync**: `argocd app set shower --revision shower-app-deploy && argocd app sync shower` to test from this branch before merging ## Test plan - [ ] Container builds successfully on nix-container-builder runner - [ ] Pod starts, migrations run, gunicorn answers on :8000 - [ ] `kubectl --context=k3s-ringtail -n shower logs deploy/shower` clean - [ ] `curl -sf https://shower.ops.eblu.me/` returns the splash page (tailnet) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/` returns 200 (pre-DNS verification) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/admin/users/` returns 403 (edge block) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/admin/login/` returns a Django login response - [ ] After DNS is up: `curl -I https://shower.eblu.me/` returns 200 with `X-Clacks-Overhead` - [ ] Grafana dashboard "Shower APM" appears and starts showing traffic - [ ] `mise run services-check` passes Reviewed-on: https://forge.eblu.me/eblume/blumeops/pulls/349
2026-05-11 13:47:18 -07:00
"$TMPDIR/venv/bin/pip" install \
--no-cache-dir \
--index-url=https://pypi.ops.eblu.me/root/pypi/+simple/ \
C1: deploy shower v1.1.0 (phases + guest memories) (#354) ## Summary Deploys `adelaide-baby-shower-app` **v1.1.0** to ringtail k3s. ### App changes (since v1.0.2) - **Four-phase `ShowerState`** replaces the boolean `locked` flag — `pre_event` → `party` → `prizes_locked` → `event_locked` — with a backfill migration that maps `locked=True → pre_event`, `locked=False → party`. - **Guest memories**: append-only photos + comments panel where guests can leave notes for the baby. Adds `GuestPhoto` + `GuestComment` models with file-extension validators and a max-size validator; new `shower.imaging` module for thumbnail generation. - **Admin + QR polish**: configurable host link, fixed "View Site" URL, guest-facing QR copy improvements, contest tweaks. Three Django migrations run automatically in the entrypoint against the SQLite PV: - `0009_shower_phase` - `0010_guest_memories` - `0011_book_description` No ConfigMap / env-var changes. The deploy uses `strategy: Recreate` with a single replica, so the old pod releases the data PVC before the new one mounts it and runs migrations. ### Container build changes The v1.1.0 tag exposed a latent issue with the Forgejo PyPI install path: - The recent commit [2d38418e](https://forge.eblu.me/eblume/blumeops/commit/2d38418e) closed the forge package leak at the Fly edge by blocking `/api/packages/*` publicly. - Forgejo's PyPI simple index returns absolute file URLs hardcoded to its public `ROOT_URL` (`forge.eblu.me`), so pip-installing from the tailnet index URL still tries to download from `forge.eblu.me` → 403. - Previous shower builds escaped this because their FOD outputs were already in the nix store; bumping to a new version forced a fresh pip run that hit the block. Fix mirrors what we already do for the sdist: both wheel and sdist are pulled via direct `fetchurl` against `forge.ops.eblu.me`, then the wheel is copied to TMPDIR under its clean filename (nix store path's hash prefix breaks pip's wheel-filename parser) and handed to pip as a local path. The forge `--extra-index-url` is no longer needed. FOD outputHash pinned to `sha256-kTNOswobtkgyQmmqbQM8XO4vvaGg57nCuuZGbNXb0NM=` from run 547. Image: `registry.ops.eblu.me/blumeops/shower:v1.1.0-444ff91-nix`. ### Adjacent finding (already handled) The ringtail `gitea-runner-nix_container_builder` systemd unit was left `inactive` after the recent `provision-ringtail` (matches the known `sshd-restart-hangs-mux` lesson — the rebuild changed the unit's PATH closure + config.yaml, systemd stopped it, then the playbook hung before the activation could restart it). Manually started; the existing memory `lesson_provision_ringtail_ssh_hang.md` was extended to mention the runner as the canary service to check after provisions. ## Test plan - [ ] `argocd app diff shower --revision shower-v1.1.0` — review the manifest change - [ ] `argocd app set shower --revision shower-v1.1.0 && argocd app sync shower` - [ ] `kubectl --context=k3s-ringtail logs -n shower deploy/shower` — confirm migrations 0009/0010/0011 applied, no errors - [ ] Hit `https://shower.ops.eblu.me/` (tailnet) — splash page renders, phase indicator visible - [ ] Hit `https://shower.ops.eblu.me/host/` — host console loads, phase dropdown shows the four states - [ ] Hit `https://shower.eblu.me/` (public via Fly) — splash page still served - [ ] After merge: `argocd app set shower --revision main && argocd app sync shower` Reviewed-on: https://forge.eblu.me/eblume/blumeops/pulls/354
2026-05-11 20:08:03 -07:00
"$TMPDIR/${showerWheel.name}" \
C1: deploy adelaide-baby-shower-app to ringtail k3s (#349) ## Summary Brings up the Adelaide / Heidi / Addie baby shower app on ringtail k3s with the public/private split that the app's hosting contract calls for: `shower.eblu.me` (public, via Fly proxy) and `shower.ops.eblu.me` (tailnet). App is consumed as a wheel from the Forgejo PyPI index — source lives at [`adelaide-baby-shower-app`](https://forge.eblu.me/eblume/adelaide-baby-shower-app). ### What's included - **ArgoCD app + manifests** under `argocd/manifests/shower/` (deployment, service, ProxyGroup ingress, ConfigMap for `DJANGO_DEBUG`/`DJANGO_ADMIN_URL`, ExternalSecret for `DJANGO_SECRET_KEY` from 1Password item `Shower (blumeops)`, NFS PV on sifaka, RWX media PVC, RWO local-path data PVC for SQLite). Recreate rollout because SQLite is single-writer. - **Public surface** (`fly/`): new `shower.eblu.me` server block proxying to `shower.ops.eblu.me`. `/admin/` returns 403 at the edge except `/admin/login/` and `/admin/logout/`, which are rate-limited via a new `shower_auth` zone. `X-Clacks-Overhead` on. GNU Terry Pratchett. - **fail2ban** filter (`shower-admin-login.conf`) matching 401/403/429 on `/admin/login/` and jail (`shower.conf`) with `maxretry=5/findtime=600/bantime=3600`. The `nginx-deny` action was generalized to take a per-jail `nginx_deny_file` so the shower has its own deny list (forge keeps using the legacy default). - **Caddy** route on indri (`shower.ops.eblu.me` → `https://shower.tail8d86e.ts.net`). - **Pulumi** Gandi CNAME `shower.eblu.me → blumeops-proxy.fly.dev.`. - **Grafana** APM dashboard `configmap-shower-apm.yaml` (request rate, error rate, failed admin login count, latency percentiles, bandwidth, access logs) mirroring `docs-apm.json` with a `host="shower.eblu.me"` filter. - **Container** `containers/shower/default.nix` — `dockerTools.buildLayeredImage` with a nixpkgs Python and a startup wrapper that creates `/app/data/.venv`, pip-installs `adelaide-baby-shower-app==1.0.0` from the forge PyPI index on first boot, runs migrations + collectstatic, and execs gunicorn. A `local_settings.py` shim pins `DATABASES.NAME`/`MEDIA_ROOT`/`STATIC_ROOT` to absolute paths so they don't end up in site-packages. - **Docs** runbook at `docs/how-to/operations/shower-app.md` linked from the apps registry, plus changelog fragments. ### Defense layers on the public surface 1. fly nginx geo+fail2ban `$shower_banned` (per-service deny list) 2. fly nginx `limit_req zone=shower_auth` (3 r/s per Fly-Client-IP) 3. django-axes (5 fails / 1h, keyed on username+ip_address) 4. edge `/admin/` block (returns 403 for anything that isn't login/logout) ## Prerequisites for the user to do (NOT in this PR) Halted on these per request — they touch shared/manual systems: - [x] **NFS share** on sifaka: `/volume1/shower`, NFS rule for ringtail RW, `chown 1000:1000` - [ ] **1Password item** `Shower (blumeops)` in the blumeops vault with a freshly minted `secret-key` field (`openssl rand -base64 48`) — do NOT reuse anything that has lived in git - [ ] **Container build**: `mise run container-build-and-release shower`, then update `images[].newTag` in `argocd/manifests/shower/kustomization.yaml` to the resulting `v1.0.0-<sha>-nix` - [x] **DNS**: `mise run dns-up` after merge - [x] **Fly cert**: `fly certs add shower.eblu.me -a blumeops-proxy` - [ ] **Caddy push**: `mise run provision-indri -- --tags caddy` - [ ] **Fly redeploy** to pick up the new nginx block + fail2ban jail: `mise run fly-deploy` - [ ] **ArgoCD sync**: `argocd app set shower --revision shower-app-deploy && argocd app sync shower` to test from this branch before merging ## Test plan - [ ] Container builds successfully on nix-container-builder runner - [ ] Pod starts, migrations run, gunicorn answers on :8000 - [ ] `kubectl --context=k3s-ringtail -n shower logs deploy/shower` clean - [ ] `curl -sf https://shower.ops.eblu.me/` returns the splash page (tailnet) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/` returns 200 (pre-DNS verification) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/admin/users/` returns 403 (edge block) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/admin/login/` returns a Django login response - [ ] After DNS is up: `curl -I https://shower.eblu.me/` returns 200 with `X-Clacks-Overhead` - [ ] Grafana dashboard "Shower APM" appears and starts showing traffic - [ ] `mise run services-check` passes Reviewed-on: https://forge.eblu.me/eblume/blumeops/pulls/349
2026-05-11 13:47:18 -07:00
gunicorn
runHook postBuild
'';
installPhase = ''
runHook preInstall
mkdir -p $out/lib/python3.14 $out/bin
cp -r "$TMPDIR/venv/lib/python3.14/site-packages" $out/lib/python3.14/site-packages
for script in "$TMPDIR/venv/bin/"*; do
[ -f "$script" ] || continue
name=$(basename "$script")
case "$name" in
python*|pip*|activate*) continue ;;
esac
cp "$script" "$out/bin/$name"
chmod +x "$out/bin/$name"
done
# --- Strip Nix store references (FOD outputs must be self-contained) ---
# The wrapper derivation below restores them via autoPatchelfHook + a
# python wrapper that points pyc-less imports at the on-image python.
# Strip bytecode entirely — pyc files embed compile-time paths.
find $out -type f -name '*.pyc' -delete
find $out -type d -name '__pycache__' -exec rm -rf {} + 2>/dev/null || true
# Dynamically discover all nix store references and strip them. We
# don't have a static list because pip pulls in stdenv via Python's
# build env (gcc-lib, libstdc++, etc.) and the closure is opaque.
{ find $out -type f -print0 \
| xargs -0 grep -aohE '/nix/store/[a-z0-9]{32}-[^/"[:space:]]+' 2>/dev/null \
|| true; } | sort -u > $TMPDIR/store-refs.txt
echo "Found $(wc -l < $TMPDIR/store-refs.txt) unique store path references to strip"
refs_args=""
while IFS= read -r ref; do
refs_args="$refs_args -t $ref"
done < $TMPDIR/store-refs.txt
if [ -n "$refs_args" ]; then
find $out -type f -exec remove-references-to $refs_args {} + 2>/dev/null || true
fi
remaining=$({ find $out -type f -print0 | xargs -0 grep -cl '/nix/store/' 2>/dev/null || true; } | wc -l)
echo "Files with remaining store references: $remaining"
runHook postInstall
'';
outputHashMode = "recursive";
outputHashAlgo = "sha256";
# Pinned dep closure — reproducible until version bumps. To recompute,
# set to pkgs.lib.fakeHash and read the failure.
outputHash = "sha256-1xx2qWAIwherklHIPXo6IOKkKHML1KUrUx6pbkMxffc=";
C1: deploy adelaide-baby-shower-app to ringtail k3s (#349) ## Summary Brings up the Adelaide / Heidi / Addie baby shower app on ringtail k3s with the public/private split that the app's hosting contract calls for: `shower.eblu.me` (public, via Fly proxy) and `shower.ops.eblu.me` (tailnet). App is consumed as a wheel from the Forgejo PyPI index — source lives at [`adelaide-baby-shower-app`](https://forge.eblu.me/eblume/adelaide-baby-shower-app). ### What's included - **ArgoCD app + manifests** under `argocd/manifests/shower/` (deployment, service, ProxyGroup ingress, ConfigMap for `DJANGO_DEBUG`/`DJANGO_ADMIN_URL`, ExternalSecret for `DJANGO_SECRET_KEY` from 1Password item `Shower (blumeops)`, NFS PV on sifaka, RWX media PVC, RWO local-path data PVC for SQLite). Recreate rollout because SQLite is single-writer. - **Public surface** (`fly/`): new `shower.eblu.me` server block proxying to `shower.ops.eblu.me`. `/admin/` returns 403 at the edge except `/admin/login/` and `/admin/logout/`, which are rate-limited via a new `shower_auth` zone. `X-Clacks-Overhead` on. GNU Terry Pratchett. - **fail2ban** filter (`shower-admin-login.conf`) matching 401/403/429 on `/admin/login/` and jail (`shower.conf`) with `maxretry=5/findtime=600/bantime=3600`. The `nginx-deny` action was generalized to take a per-jail `nginx_deny_file` so the shower has its own deny list (forge keeps using the legacy default). - **Caddy** route on indri (`shower.ops.eblu.me` → `https://shower.tail8d86e.ts.net`). - **Pulumi** Gandi CNAME `shower.eblu.me → blumeops-proxy.fly.dev.`. - **Grafana** APM dashboard `configmap-shower-apm.yaml` (request rate, error rate, failed admin login count, latency percentiles, bandwidth, access logs) mirroring `docs-apm.json` with a `host="shower.eblu.me"` filter. - **Container** `containers/shower/default.nix` — `dockerTools.buildLayeredImage` with a nixpkgs Python and a startup wrapper that creates `/app/data/.venv`, pip-installs `adelaide-baby-shower-app==1.0.0` from the forge PyPI index on first boot, runs migrations + collectstatic, and execs gunicorn. A `local_settings.py` shim pins `DATABASES.NAME`/`MEDIA_ROOT`/`STATIC_ROOT` to absolute paths so they don't end up in site-packages. - **Docs** runbook at `docs/how-to/operations/shower-app.md` linked from the apps registry, plus changelog fragments. ### Defense layers on the public surface 1. fly nginx geo+fail2ban `$shower_banned` (per-service deny list) 2. fly nginx `limit_req zone=shower_auth` (3 r/s per Fly-Client-IP) 3. django-axes (5 fails / 1h, keyed on username+ip_address) 4. edge `/admin/` block (returns 403 for anything that isn't login/logout) ## Prerequisites for the user to do (NOT in this PR) Halted on these per request — they touch shared/manual systems: - [x] **NFS share** on sifaka: `/volume1/shower`, NFS rule for ringtail RW, `chown 1000:1000` - [ ] **1Password item** `Shower (blumeops)` in the blumeops vault with a freshly minted `secret-key` field (`openssl rand -base64 48`) — do NOT reuse anything that has lived in git - [ ] **Container build**: `mise run container-build-and-release shower`, then update `images[].newTag` in `argocd/manifests/shower/kustomization.yaml` to the resulting `v1.0.0-<sha>-nix` - [x] **DNS**: `mise run dns-up` after merge - [x] **Fly cert**: `fly certs add shower.eblu.me -a blumeops-proxy` - [ ] **Caddy push**: `mise run provision-indri -- --tags caddy` - [ ] **Fly redeploy** to pick up the new nginx block + fail2ban jail: `mise run fly-deploy` - [ ] **ArgoCD sync**: `argocd app set shower --revision shower-app-deploy && argocd app sync shower` to test from this branch before merging ## Test plan - [ ] Container builds successfully on nix-container-builder runner - [ ] Pod starts, migrations run, gunicorn answers on :8000 - [ ] `kubectl --context=k3s-ringtail -n shower logs deploy/shower` clean - [ ] `curl -sf https://shower.ops.eblu.me/` returns the splash page (tailnet) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/` returns 200 (pre-DNS verification) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/admin/users/` returns 403 (edge block) - [ ] `curl -I -H "Host: shower.eblu.me" https://blumeops-proxy.fly.dev/admin/login/` returns a Django login response - [ ] After DNS is up: `curl -I https://shower.eblu.me/` returns 200 with `X-Clacks-Overhead` - [ ] Grafana dashboard "Shower APM" appears and starts showing traffic - [ ] `mise run services-check` passes Reviewed-on: https://forge.eblu.me/eblume/blumeops/pulls/349
2026-05-11 13:47:18 -07:00
dontFixup = true;
};
# Non-FOD wrapper: re-applies RPATHs to pre-built .so files (pillow,
# scipy) so they find libstdc++ / libz / etc. at runtime. autoPatchelfHook
# discovers needed libraries from buildInputs.
pyDeps = pkgs.stdenv.mkDerivation {
pname = "shower-python-deps";
inherit version;
dontUnpack = true;
nativeBuildInputs = [ pkgs.autoPatchelfHook ];
buildInputs = with pkgs; [
python
stdenv.cc.cc.lib # libstdc++, libgcc_s
zlib
libjpeg
libwebp
libtiff
openjpeg
lcms2
freetype
];
installPhase = ''
cp -r ${pyDepsFOD} $out
chmod -R u+w $out
'';
};
sitePackages = "${pyDeps}/lib/python3.14/site-packages";
# Settings shim — config/settings.py's `BASE_DIR = parent.parent` would
# otherwise resolve to site-packages, scattering db.sqlite3 / media /
# staticfiles into the venv. Pin them to /app/{data,media,data/staticfiles}.
localSettings = pkgs.writeText "local_settings.py" ''
from pathlib import Path
from config.settings import * # noqa: F401,F403
DATABASES["default"]["NAME"] = "/app/data/db.sqlite3"
MEDIA_ROOT = "/app/media"
STATIC_ROOT = "/app/data/staticfiles"
# /app/static comes from the repo-root static/ subtree of the sdist
# (see default.nix staticAssets). Added because the wheel doesn't
# ship vendored Sortable/cropper assets.
STATICFILES_DIRS = [Path("/app/static")]
'';
# PYTHONPATH, DJANGO_SETTINGS_MODULE, PATH, and HOME live in the image's
# `Env` block below — that way `kubectl exec deploy/shower -- python -m
# django <subcommand>` Just Works without an inline `env` ceremony.
# The entrypoint just changes directory and runs the boot sequence.
entrypoint = pkgs.writeShellScript "shower-entrypoint" ''
set -eu
cd /app
mkdir -p /app/data /app/media
echo "shower: running migrations"
python -m django migrate --noinput
echo "shower: collecting static files"
python -m django collectstatic --noinput --clear
echo "shower: starting gunicorn"
exec gunicorn \
--bind 0.0.0.0:8000 \
--workers 2 \
--forwarded-allow-ips='*' \
config.wsgi:application
'';
in
pkgs.dockerTools.buildLayeredImage {
name = "blumeops/shower";
contents = [
python
pyDeps
pkgs.cacert
pkgs.tzdata
pkgs.bashInteractive
pkgs.coreutils
];
extraCommands = ''
mkdir -p app/data app/media tmp
chmod 1777 tmp
cp ${localSettings} app/local_settings.py
cp -r ${staticAssets} app/static
chmod -R u+w app/static
'';
fakeRootCommands = ''
chown -R 1000:1000 app
'';
enableFakechroot = true;
config = {
Entrypoint = [ "${entrypoint}" ];
Env = [
"SSL_CERT_FILE=${pkgs.cacert}/etc/ssl/certs/ca-bundle.crt"
"TZDIR=${pkgs.tzdata}/share/zoneinfo"
"TZ=America/Los_Angeles"
"TMPDIR=/tmp"
"LANG=C.UTF-8"
"LC_ALL=C.UTF-8"
"PYTHONDONTWRITEBYTECODE=1"
"HOME=/app/data"
"PATH=${pyDeps}/bin:${python}/bin:/bin"
# /app first so local_settings.py is importable; sitePackages second so
# django, gunicorn, etc. resolve. Inherited by entrypoint + any
# `kubectl exec` so manual django subcommands work without ceremony.
"PYTHONPATH=/app:${sitePackages}"
"DJANGO_SETTINGS_MODULE=local_settings"
];
ExposedPorts = {
"8000/tcp" = { };
};
User = "1000";
WorkingDir = "/app";
};
}