tidy up notes

This commit is contained in:
John Gatward committed 2026-10-27 12:21:10 +00:00
1 parent 8e21e897a5
commit b17e750121
15 files changed
+669 -2078

No files matched your search

+63 -220
View File
@@ -4,213 +4,90 @@ id: home-server.authentication
type: reference
scope: [jupiter, authentik, oidc, forward-auth, per-user-access, secrets]
sensitivity: private-infrastructure
last_reviewed: "2026-10-26"
last_reviewed: "2026-10-27"
sources:
- kind: owner-report
reference: "Public qbittorrent.umbra.mom UI working through Mercury/Traefik/Authentik with native qBittorrent login retained, 2026-10-26; does not establish broader per-user migration completion"
- kind: owner-report
reference: "Requested removal of custom service-access groups, manual per-user application access and retained Jay administration, 2026-10-26; repository prepared, deployment unconfirmed"
- kind: owner-report
reference: "Requested Authentik-only Homarr login and read-only Docker socket proxy; deployment not confirmed, 2026-10-18"
- kind: owner-report
reference: "Homarr Authentik login created a separate account; requested linking to existing jay account, 2026-10-18"
reference: "Authentik deployed/Authelia retired 2026-10-16; Homarr login 2026-10-18; per-user policy and account linking still unconfirmed 2026-10-26"
- kind: repository
repository: jupiter-stacks
revision: 026cfa58b601165342b278401183c5fed5a50db2
paths:
- stacks/homarr/docker-compose.yml
- stacks/homarr/.env.example
- stacks/authentik/docker-compose.yml
- stacks/authentik/blueprints/jupiter.yaml
- stacks/authentik/.env.example
- kind: owner-report
reference: "Authentik fully deployed, all listed app integrations complete and Authelia retired; OIDC and three-group motivation, 2026-10-16"
- kind: owner-report
reference: "Six-directory Backrest plan including Authentik and Wardrowbe; deployed Authentik container names confirmed, 2026-10-16"
- kind: owner-report
reference: "All stack environment values/secrets managed in Portainer; GitHub access uses PAT, 2026-10-06"
- kind: owner-report
reference: "Independent credential recovery unchanged: only restic key has independent custody, 2026-10-16"
- kind: repository
repository: jupiter-stacks
revision: d09df456ad331b8beb9a40615114ea8aae2dced5
paths:
- stacks/authentik/docker-compose.yml
- stacks/authentik/blueprints/jupiter.yaml
- stacks/authentik/.env.example
- stacks/beszel/docker-compose.yml
- stacks/dozzle/docker-compose.yml
- stacks/gitea/docker-compose.yml
- stacks/immich/docker-compose.yml
- stacks/paperless-ngx/docker-compose.yml
- stacks/portainer/portainer_start.sh
- stacks/wardrowbe/docker-compose.yml
- stacks/backrest/docker-compose.yml
- stacks/notes/docker-compose.yml
- stacks/prowlarr/docker-compose.yml
- stacks/radarr/docker-compose.yml
- stacks/sonarr/docker-compose.yml
related: [home-server.reference, home-server.networking, home-server.portainer, home-server.backups, home-server.deployment]
update_triggers: [access-policy-change, authentication-change, secret-rotation, middleware-change]
unknowns:
- deployed-per-user-bindings-and-jay-superuser-membership
- per-application-local-login-and-account-linking-settings
- retired-authelia-data-retention
revision: 12fb1b96e01fc062fdce4c216f11eb5d7ce716ed
paths: [stacks/authentik/docker-compose.yml, stacks/authentik/blueprints/jupiter.yaml, stacks/authentik/.env.example, stacks/homarr/docker-compose.yml]
related: [home-server.networking, home-server.deployment, home-server.backups, home-server.portainer]
update_triggers: [access-policy-change, authentication-change, secret-rotation]
unknowns: [deployed-per-user-bindings-and-jay-superuser-membership, per-application-local-login-and-account-linking, retired-authelia-data-retention]
---
# Authentication
Authentik replaces Authelia. The original setup used native OIDC and forward-auth
with three access groups. On 2026-10-26 the owner requested direct per-user
service access instead; the repository now prepares that migration, with `jay`
allowed into every configured application. Deployment is unconfirmed.
The owner reports the earlier Authentik migration
fully deployed as of 2026-10-16 and Authelia retired. Homarr OIDC login was
reported working on 2026-10-18; linking to the existing account is configured
in the repository but has not been confirmed successful.
Configuration details below are repository evidence, not live inspection.
| Component | Configuration |
| --- | --- |
| Portal | `https://auth.umbra.mom` |
| Identity service | Authentik server + worker; image `2026.8.4` |
| Database | PostgreSQL 16; database/user `authentik` |
| Configuration | `stacks/authentik/blueprints/jupiter.yaml`; worker-mounted custom blueprint |
| Application MFA | `jupiter-app-authorization`; TOTP or WebAuthn; prompts for setup if none configured |
| Recent MFA | Five-minute reuse threshold avoids an immediate duplicate challenge |
| Host setup access | Loopback `127.0.0.1:9000` -> server `9000` |
- **Deployed:** Authentik replaces Authelia; portal `https://auth.umbra.mom`.
- **Repository intent, deployment unconfirmed:** direct per-user access and Homarr account linking.
- Backend: server + worker + PostgreSQL 16 (`authentik` DB/user).
- Blueprint: `stacks/authentik/blueprints/jupiter.yaml`; host copy must be applied separately.
- MFA: `jupiter-app-authorization`; TOTP/WebAuthn; setup if absent; five-minute reuse.
- Proxy transport/trust: [Networking](networking.md#listener-and-trust-boundaries).
## Per-user application access
Every one of the 15 configured applications has a direct binding for the
**existing** Authentik user `jay`, at order `0`. Applications use
`policy_engine_mode: any`: Jay **or** another explicitly added user may pass.
Unassigned users are denied, and the shared application-MFA flow remains.
The blueprint does not create users or change passwords, MFA devices or
application-native administrator roles.
Add other users in the application's **Policy / Group / User Bindings** tab,
using a **User** binding with order `10` or higher, enabled and not negated.
Keep ANY mode and retain Jay's order-0 binding. Manual bindings at other orders
are not removed by blueprint reapplication. Admission does not itself grant
administrator privileges inside a service.
The blueprint deletes only the custom `admins`, `development` and `clients`
groups after replacing their bindings; their memberships and group-based grants
are removed. **Retain Jay's built-in superuser group membership**, normally
`authentik Admins`: Authentik derives administrator status from a superuser
group, so removing every group would lock out administration.
Before applying, back up Authentik, confirm Jay is still a superuser, and redeploy
Trek with its username-based admin mapping. Copy the updated blueprint into the
host's `/data/authentik/blueprints/jupiter.yaml` and apply it, then individually
assign the other users you want to retain. See the
[access-management runbook](https://github.com/jayo60013/jupiter-stacks/blob/main/stacks/authentik/README.md)
for rollout and rollback. Existing application sessions/API keys need separate
revocation when removing access.
- All 15 applications bind existing user `jay` at order `0`; `policy_engine_mode: any`.
- Add others under **Policy / Group / User Bindings**: enabled, non-negated **User** binding, order `10+`. Unassigned users are denied; manual higher-order bindings survive reapplication.
- Blueprint replaces custom `admins`, `development`, `clients` groups; does not create users or change passwords/MFA/app-native roles.
- **Keep Jay in the built-in superuser group** (`authentik Admins`); application admission is not administration.
- Before applying: back up Authentik, verify Jay's superuser membership, redeploy Trek's username-based admin mapping, copy/apply the host blueprint, assign retained users.
- Revoke existing application sessions/API keys separately when removing access.
## Native OIDC
Each client uses a confidential authorization-code provider, strict redirect URIs,
per-provider issuer and the shared application-MFA authorization flow.
Issuer pattern: `https://auth.umbra.mom/application/o/<client-id>/`.
- Confidential authorization-code clients; strict callbacks; shared MFA flow.
- Issuer: `https://auth.umbra.mom/application/o/<client-id>/`.
| Client ID | Application hostname | Integration configuration |
| Client ID | Host under `umbra.mom` | Configuration |
| --- | --- | --- |
| `beszel` | `beszel.umbra.mom` | Database/UI provider; password-auth disable flag in Compose |
| `dozzle` | `logs.umbra.mom` | Compose OIDC settings; no forward-auth middleware |
| `gitea` | `gitea.umbra.mom` | Database/UI authentication source; Compose requires login for account linking and disables automatic registration |
| `homarr` | `umbra.mom` | Compose OIDC settings; Authentik-only login configured; verified-email linking to existing password accounts enabled |
| `immich` | `photos.umbra.mom` | Database/UI OAuth settings; web and mobile callbacks registered |
| `paperless` | `paperless.umbra.mom` | Compose OpenID Connect/allauth configuration; PKCE enabled; signups disabled |
| `portainer` | `portainer.umbra.mom`, `portainer.local.umbra.mom` | Database/UI OAuth settings; both callback hosts registered |
| `trek` | `holidays.umbra.mom` | Compose OIDC settings; `preferred_username == jay` grants the app admin role |
| `wardrowbe` | `wardrobe.umbra.mom` | Compose OIDC settings for frontend and backend/workers |
| `beszel` | `beszel` | Database/UI provider; password-disable flag in Compose |
| `dozzle` | `logs` | Compose OIDC; `dozzle_roles: ["none"]`; read-only Docker proxy |
| `gitea` | `gitea` | Database/UI source; login required for linking; auto-registration disabled |
| `homarr` | Apex | Compose OIDC; verified-email linking |
| `immich` | `photos` | Database/UI OAuth; web + mobile callbacks |
| `paperless` | `paperless` | Compose allauth/OIDC; PKCE; signups disabled |
| `portainer` | `portainer`, `portainer.local` | Database/UI OAuth; both callback hosts |
| `trek` | `holidays` | Compose; `preferred_username == jay` grants app admin |
| `wardrowbe` | `wardrobe` | Compose frontend/backend/workers |
Dozzle's custom profile mapping emits `dozzle_roles: ["none"]` for any user
admitted by its direct application bindings. Its Docker proxy stays read-only.
The email mapping asserts `email_verified` only when the user has
an email and their `email_verified` attribute is explicitly true.
OIDC integration does not prove local password login is disabled or application
roles are synchronized. Beszel and Paperless retain migration-safe local-login
defaults in Compose; deployed overrides and account-linking details were not
supplied. Application permissions and existing sessions remain app-managed.
- Beszel/Paperless retain migration-safe local-login defaults; deployed overrides unconfirmed.
- OIDC does not synchronise native app roles or prove local login disabled.
- Email mapping emits `email_verified` only for an existing email with explicit boolean `true`.
### Homarr
Homarr uses native OIDC, not a Traefik forward-auth gate. Its Authentik provider
is `jupiter-homarr`, with the shared application-MFA flow.
| Setting | Value |
| --- | --- |
| Login providers | `AUTH_PROVIDERS=oidc` (Authentik only) |
| Issuer | `https://auth.umbra.mom/application/o/homarr/` |
| Client ID / display name | `homarr` / `Authentik` |
| Callback | `https://umbra.mom/api/auth/callback/oidc` (strict match) |
| Requested scopes | `openid email profile` |
| Client secret | `OIDC_HOMARR_SECRET` in both the Homarr and Authentik stack environments |
| Existing-account linking | `AUTH_OIDC_ENABLE_DANGEROUS_CREDENTIALS_LINKING=true` |
| Provider / login | `jupiter-homarr`; `AUTH_PROVIDERS=oidc` |
| Callback | `https://umbra.mom/api/auth/callback/oidc` |
| Scopes | `openid email profile` |
| Shared secret | `OIDC_HOMARR_SECRET` in Homarr + Authentik |
| Linking | `AUTH_OIDC_ENABLE_DANGEROUS_CREDENTIALS_LINKING=true` |
An OIDC login initially created a separate user instead of signing into the
existing `jay` password account. Linking is now enabled in Compose; it matches
email addresses, not usernames, and preserves the existing account's username,
password credentials, permissions, boards, settings and API keys. Password
credentials are retained but cannot be used while only OIDC is enabled.
- Linking matches **email**, not username; applies to all matching password accounts.
- Preserve `SECRET_ENCRYPTION_KEY`; password credentials remain stored but unusable with OIDC-only login.
To complete linking:
1. Back up `/data/homarr/appdata`; ensure existing `jay` email matches the intended identity.
2. Independently verify ownership, then set that user's `email_verified` to boolean `true`; never mark all users verified.
3. Preserve/transfer duplicate-account data, then delete **only the duplicate OIDC account**, not `jay`; an existing link otherwise wins.
4. Confirm login reaches `jay`, admin permissions and existing boards before disabling credentials. Temporarily use `AUTH_PROVIDERS=credentials,oidc` if needed.
1. Before deploying Authentik-only login, back up `/data/homarr/appdata`, then
sign in with the local administrator account and ensure `jay`'s email exactly
matches the intended Authentik user's.
2. Independently verify ownership of that email before setting the Authentik
user's `email_verified` attribute to boolean `true`, not string `"true"`.
The existing email scope mapping requires this before issuing
`email_verified: true`; do not mark every user's email verified.
3. Preserve or transfer anything owned by the duplicate OIDC account, then
delete only that duplicate through Homarr's user management. Do not delete
`jay`. An existing OIDC link otherwise continues using the duplicate.
4. Confirm Authentik login links to `jay` with administrator access and the
existing boards and permissions before disabling password login. If linking
still needs setup, temporarily use `AUTH_PROVIDERS=credentials,oidc` and
redeploy. Once working, pull/redeploy with `AUTH_PROVIDERS=oidc`.
The linking setting applies to all matching password accounts, not just `jay`.
Only trusted administrators should mark Authentik emails verified. Password login
is disabled in the current Compose configuration; successful deployment and
account linking have not been confirmed. For recovery if Authentik is unavailable,
temporarily restore `AUTH_PROVIDERS=credentials,oidc` and redeploy Homarr.
That recovery setting bypasses Authentik MFA and application-access policies
for password logins. Keep `SECRET_ENCRYPTION_KEY` unchanged.
Homarr also has a read-only Docker socket proxy on an internal `docker-api`
network. Only the proxy mounts `/var/run/docker.sock`; its API port is not
published or connected to Traefik. Homarr's `Jupiter` endpoint permits inventory
and logs, while the proxy blocks writes (`POST=0`). Container lifecycle controls
are disabled. Redeploy the whole stack to create the proxy alongside Homarr.
- Credential fallback bypasses Authentik MFA/access policy for password logins.
- Internal `docker-api` proxy: only proxy mounts socket; no published API/Traefik attachment; inventory/logs allowed, `POST=0`, lifecycle controls disabled. Redeploy the whole stack.
## Forward-auth
Backrest (`backups.umbra.mom`), Notes, Prowlarr, Radarr and Sonarr retain
proxy-based protection. Their routers attach `authentik@docker`:
- Protected apps: Backrest (`backups`), Notes, Prowlarr, Radarr, Sonarr, qBittorrent.
- Router middleware: `authentik@docker`; `trustForwardHeader=true`.
- Endpoint: `http://authentik-server:9000/outpost.goauthentik.io/auth/traefik`.
- Six embedded `forward_single` providers; explicit external hosts, empty `skip_path_regex`.
- Priority **1000** callback router: `/outpost.goauthentik.io/` goes directly to Authentik, bypassing forward-auth.
- qBittorrent keeps native login; internal Arr API traffic bypasses this browser gate.
- Only attached routers are protected; published ports, SSH and API tokens are separate boundaries.
```text
Traefik -> http://authentik-server:9000/outpost.goauthentik.io/auth/traefik
```
The blueprint now includes a sixth `forward_single` provider for
`qbittorrent.umbra.mom`, with a direct `jay` user binding and the shared MFA flow.
The owner confirms its public UI and both authentication layers are working
on 2026-10-26. This confirmation is specific to qBittorrent, not proof that the
broader per-user access migration is complete. qBittorrent retains its own password login, and internal Arr API
connections do not pass through Authentik. See [Local downloads](downloads.md).
The embedded-outpost blueprint lists six `forward_single` providers, each with an
explicit external host and empty `skip_path_regex`. A priority-1000 router
on those hosts sends `/outpost.goauthentik.io/` directly to Authentik,
bypassing forward-auth so callbacks can complete.
`trustForwardHeader=true`; forwarded response headers:
Forwarded response headers:
```text
X-authentik-username, X-authentik-groups, X-authentik-entitlements,
@@ -219,50 +96,16 @@ X-authentik-meta-jwks, X-authentik-meta-outpost, X-authentik-meta-provider,
X-authentik-meta-app, X-authentik-meta-version
```
Forward-auth applies only to attached routers. Authentik is not a global gate
for every service, directly published port, SSH connection or API token.
## State / secrets
| Host path | Container use |
| --- | --- |
| `/data/authentik/postgres` | PostgreSQL `/var/lib/postgresql/data`; identity, policy and MFA state |
| `/data/authentik/data` | Server + worker `/data`; persistent files |
| `/data/authentik/blueprints` | Worker `/blueprints/custom:ro`; host copy of repository blueprint |
| `/data/authentik/postgres` | PostgreSQL `/var/lib/postgresql/data` |
| `/data/authentik/data` | Server + worker `/data` |
| `/data/authentik/blueprints` | Worker `/blueprints/custom:ro` |
Stack variable overrides/secrets are managed in Portainer; Compose retains
non-secret configuration. Required Authentik stack secrets:
```text
AUTHENTIK_SECRET_KEY
AUTHENTIK_DB_PASSWORD
OIDC_BESZEL_SECRET
OIDC_DOZZLE_SECRET
OIDC_GITEA_SECRET
OIDC_HOMARR_SECRET
OIDC_IMMICH_SECRET
OIDC_PAPERLESS_SECRET
OIDC_PORTAINER_SECRET
OIDC_WARDROWBE_SECRET
```
Client secrets are independently generated plaintext credentials, not old
Authelia password hashes. The worker passes them to blueprint `!Env` entries.
Keep each provider secret matched to its application configuration:
Homarr uses `OIDC_HOMARR_SECRET`, Dozzle uses `DOZZLE_OIDC_CLIENT_SECRET`, Paperless
`PAPERLESS_OIDC_CLIENT_SECRET`, and Wardrowbe `WARDROWBE_OIDC_CLIENT_SECRET`.
Other clients store provider configuration in their databases/UIs.
Public Traefik routing defaults off (`AUTHENTIK_PUBLIC_ROUTING=false`) for
private initial setup; enable only after securing the administrator account.
The owner reports public deployment complete; this default is not deployed state.
See [deployment](deployment.md) for the configuration surfaces and order.
`/data/authentik` and [Portainer state](portainer.md) are included in
[Backrest](backups.md). Backups stop Authentik server/worker and PostgreSQL;
new OIDC logins and forward-auth access are unavailable during that stop.
Recovery remains untested, with no independent admin/MFA or stack-secret
recovery arrangement; only the restic key has independent custody.
Retention/deletion of retired `/data/authelia` data is unconfirmed.
GitHub repository access uses a PAT; host SSH keys and tunnel credential files
are separate from stack environment variables.
- `authentik-data`: internal DB/server/worker network; worker also uses `authentik-egress`, server uses `traefik`. Internal DB network does not imply full egress isolation.
- Secret inventory: `stacks/authentik/.env.example`; values in Portainer, consumed through `!Env`. Do not reuse Authelia hashes as client secrets.
- Match provider/application credentials: Homarr `OIDC_HOMARR_SECRET`; Dozzle `DOZZLE_OIDC_CLIENT_SECRET`; Paperless `PAPERLESS_OIDC_CLIENT_SECRET`; Wardrowbe `WARDROWBE_OIDC_CLIENT_SECRET`.
- Initial setup: `AUTHENTIK_PUBLIC_ROUTING=false`, SSH-forwarded `127.0.0.1:9000`; secure admin/MFA before public routing, prove OIDC before disabling local logins.
- Backup coverage, login downtime and credential recovery: [Backups](backups.md).
+42 -118
View File
@@ -4,156 +4,80 @@ id: home-server.backups
type: reference
scope: [jupiter, backrest, restic, google-drive, recovery]
sensitivity: private-infrastructure
last_reviewed: "2026-10-26"
last_reviewed: "2026-10-27"
sources:
- kind: owner-report
reference: "Live AirVPN downloads; qBittorrent state/defaults and AirVPN WireGuard secrets not yet added to encrypted off-host backups, 2026-10-26"
- kind: owner-report
reference: "Backrest plan JSON, destination, key custody and no restore test, 2026-10-06"
- kind: owner-report
reference: "Updated four-service Backrest plan and deployed Portainer bind mount/container name, 2026-10-06"
- kind: owner-report
reference: "Only restic key has independent custody; no independent credential recovery or backup/restore evidence, 2026-10-06"
- kind: owner-report
reference: "Updated six-directory Backrest plan including Authentik and Wardrowbe, deployed Authentik hook container names confirmed; independent recovery custody unchanged, 2026-10-16"
- kind: owner-report
reference: "Backrest notifications received in Discord server Jupiter, channel #backrest-alerts, 2026-10-23"
reference: "Plan/hooks/key custody 2026-10-06; six-directory plan and Authentik names 2026-10-16; downloads not backed up 2026-10-26; no restore evidence"
- kind: repository
repository: jupiter-stacks
revision: d09df456ad331b8beb9a40615114ea8aae2dced5
paths:
- stacks/backrest/docker-compose.yml
- stacks/gitea/docker-compose.yml
- stacks/paperless-ngx/docker-compose.yml
- stacks/vaultwarden/docker-compose.yml
- stacks/authentik/docker-compose.yml
- stacks/wardrowbe/docker-compose.yml
related: [home-server.reference, home-server.downloads, home-server.storage, home-server.portainer, home-server.authentication, home-server.operations]
paths: [stacks/backrest/docker-compose.yml, stacks/authentik/docker-compose.yml, stacks/wardrowbe/docker-compose.yml]
related: [home-server.storage, home-server.downloads, home-server.authentication, home-server.operations]
update_triggers: [backup-plan-change, destination-change, hook-change, credential-change, restore-test]
unknowns:
- last-successful-backup-and-repository-check
- google-drive-repository-path
- measured-rpo-and-rto
unknowns: [last-successful-backup-and-check, google-drive-repository-path, measured-rpo-and-rto]
---
# Backups
| Setting | Value |
| --- | --- |
| Engine | Backrest / restic; plan `jupiter-backup` |
| Repository | `jayo60013-gdrive`; Google Drive; account `jayo60013` |
| Schedule | `0 3 * * *`; UTC; daily 03:00 |
| Retention | `policyKeepLastN: 4`; last four snapshots, not four days |
| Encryption key | iCloud password app; no key stored here |
| Independent credential recovery | None beyond the restic encryption key |
| Backup / integrity evidence | None supplied |
| Restore test | Never performed |
## Notifications
Backrest notifications are delivered to **Jupiter** (Discord server) in
`#backrest-alerts`; receipt confirmed by the owner on 2026-10-23.
Exact enabled hook conditions have not been recorded. Delivery alone does not
confirm backup integrity, successful application restarts or detection of a
backup that never runs. See [Observability](operations.md#observability) for
the shared notification destinations and remaining coverage gaps.
| Engine / plan | Backrest + restic / `jupiter-backup` |
| Repository | `jayo60013-gdrive`; Google Drive account `jayo60013` |
| Schedule | `0 3 * * *`, UTC |
| Retention | `policyKeepLastN: 4`: four **snapshots**, not days |
| Key custody | iCloud password app; only independently held recovery secret |
| Evidence | No successful backup/integrity evidence supplied; restore never exercised |
## Coverage
Backrest bind-mounts host `/data` at container `/userdata:ro`.
| Host source | Plan path |
| --- | --- |
| `/data/authentik` | `/userdata/authentik` |
| `/data/gitea` | `/userdata/gitea` |
| `/data/paperless-ngx` | `/userdata/paperless-ngx` |
| `/data/portainer` | `/userdata/portainer` |
| `/data/vaultwarden` | `/userdata/vaultwarden` |
| `/data/wardrowbe` | `/userdata/wardrowbe` |
Case-insensitive exclusions: `/userdata/gitea/gitea/packages`,
`/userdata/wardrowbe/redis`.
`excludes: []`; CLI `--skip-if-unchanged` supplied despite
plan `skipIfUnchanged: false`. Unchanged runs may create no snapshot.
- Backrest sees `/data` as `/userdata:ro`; **only** these six directories are selected:
`authentik`, `gitea`, `paperless-ngx`, `portainer`, `vaultwarden`, `wardrowbe`.
- Case-insensitive exclusions: `/userdata/gitea/gitea/packages`, `/userdata/wardrowbe/redis`.
- `excludes: []`; CLI `--skip-if-unchanged` despite plan `skipIfUnchanged: false`; unchanged runs may create no snapshot.
### Local-downloads recovery gap
The owner confirms on 2026-10-26 that the live [AirVPN download stack](downloads.md)
has **not** yet been added to encrypted off-host backups. Missing recovery state:
`/data/qbittorrent/config` (settings/torrent state),
`/data/qbittorrent/defaults` (defaults and startup/resolver files), and
`/data/gluetun/secrets` (WireGuard private/preshared keys). The reserved peer
port, tunnel address and server-country settings also need recoverable
deployment configuration; no secret values belong in these notes.
Torrent payloads and media libraries remain outside the six-directory plan.
Hardlinks are not backups. Ultra.cc is retired and is no longer a remote
source to assume during recovery. Removing `rclone-seedbox.service` does not
remove Backrest's separate Google Drive/rclone dependency.
- [qBittorrent config/defaults and AirVPN keys](downloads.md#state-and-redeployment) are **not backed up** as of 2026-10-26.
- Keep address/country/reserved-port configuration recoverable alongside encrypted keys; never put key values here.
- Torrent payloads/media are also outside the plan. Retired Ultra.cc is not a recovery source.
- Backrest's Google Drive/rclone dependency is separate from the removed seedbox mount.
## Consistency hooks
Stop timeout `120s`; chains use `&&`. Errors: `ON_ERROR_FATAL`,
except Portainer, Wardrowbe and Authentik restart hooks (`ON_ERROR_IGNORE`).
Rows below group the supplied hooks by service; the plan lists Wardrowbe's
hooks before Authentik's.
- Cold filesystem copies, not DB dumps; stop timeout **120s**, ordered chains use `&&`.
- Default failure: `ON_ERROR_FATAL`. Portainer/Wardrowbe/Authentik **restart** failures: `ON_ERROR_IGNORE`.
| Event | Ordered commands / containers |
| --- | --- |
| `SNAPSHOT_START` | Stop `paperless` -> `paperless_db` |
| `SNAPSHOT_START` | Stop `gitea_runner` -> `gitea` -> `gitea_db` |
| `SNAPSHOT_START` | Stop `vaultwarden` |
| `SNAPSHOT_START` | Stop `portainer` |
| `SNAPSHOT_END`, `SNAPSHOT_ERROR` | Start `paperless_db` -> `paperless` |
| `SNAPSHOT_END`, `SNAPSHOT_ERROR` | Start `gitea_db` -> `gitea` -> `gitea_runner` |
| `SNAPSHOT_END`, `ANY_ERROR` | Start `vaultwarden` |
| `SNAPSHOT_END`, `SNAPSHOT_ERROR` | Start `portainer`; ignore hook failure |
| `SNAPSHOT_START` | Stop `wardrowbe-frontend`, `wardrowbe-backend`, `wardrowbe-worker`, `wardrowbe-image-worker` in one command -> stop `wardrowbe-postgres`, `wardrowbe-redis` |
| `SNAPSHOT_END`, `SNAPSHOT_ERROR` | Start `wardrowbe-postgres`, `wardrowbe-redis` -> wait for PostgreSQL and Redis -> start backend, worker, image-worker, frontend; ignore hook failure |
| `SNAPSHOT_START` | Stop `authentik-server`, `authentik-worker` in one command -> stop `authentik-postgresql` |
| `SNAPSHOT_END`, `SNAPSHOT_ERROR` | Start `authentik-postgresql` -> wait for PostgreSQL -> start `authentik-server`, `authentik-worker`; ignore hook failure |
| Service | `SNAPSHOT_START` stop order | Restart order / event |
| --- | --- | --- |
| Paperless | `paperless` -> `paperless_db` | DB -> app; `SNAPSHOT_END`, `SNAPSHOT_ERROR` |
| Gitea | `gitea_runner` -> `gitea` -> `gitea_db` | DB -> app -> runner; end/error |
| Vaultwarden | `vaultwarden` | Start; `SNAPSHOT_END`, `ANY_ERROR` |
| Portainer | `portainer` | Start; end/error; failure ignored |
| Wardrowbe | frontend/backend/worker/image-worker together -> postgres/redis | DB/cache -> readiness waits -> apps; end/error; failure ignored |
| Authentik | `authentik-server`, `authentik-worker` -> `authentik-postgresql` | DB -> readiness wait -> server/worker; end/error; failure ignored |
Cold filesystem copies of applications/databases; no database dumps supplied.
Paperless broker stays running. Paperless/Gitea restart chains use `docker start`
without waiting for database readiness.
Wardrowbe waits inside the containers using `pg_isready -U wardrowbe -d wardrowbe`
then `redis-cli ping`; Authentik uses `pg_isready -U authentik -d authentik`.
These are `until` loops with two-second sleeps and no explicit deadline in the
commands. Starting multiple application containers in one command does not
establish application readiness or dependency ordering.
The owner confirms the three Authentik hook names match deployed containers.
Repository Compose explicitly names only `authentik-server`; worker/database
names must be preserved or the hooks updated when recreating the stack.
Hook failure can leave services stopped; Portainer, Wardrowbe and Authentik
restart failures do not fail the backup operation. Authentik downtime also
interrupts new OIDC logins and proxy-protected access, including the Backrest UI.
Application readiness and recovery are not demonstrated by the plan.
- Wardrowbe names use `wardrowbe-` prefix; hooks precede Authentik's in the supplied plan.
- Waits: `pg_isready -U wardrowbe -d wardrowbe`, `redis-cli ping`; Authentik `pg_isready -U authentik -d authentik`; 2s polling, no explicit deadline.
- Paperless broker stays running; Paperless/Gitea use `docker start` without DB readiness waits.
- Preserve deployed Authentik worker/DB names or update hooks; only server has explicit Compose name.
- Authentik stops interrupt new OIDC/proxy logins, including Backrest access.
- **Backup completion does not prove service recovery**; ignored restart errors can leave apps down. [Notifications](operations.md#observability) do not prove integrity/readiness.
## Backrest state
| Host path | Container path |
| --- | --- |
| `/data/backrest/data` | `/data` |
| `/data/backrest/config` | `/config`; plan/repository configuration `config.json` |
| `/data/backrest/config` | `/config`; `config.json` plan/repository settings |
| `/data/backrest/rclone` | `/root/.config/rclone` |
| `/home/jay/data/backrest/cache` | `/cache` |
| `/data/backrest/restore` | `/restore` |
| `/var/run/docker.sock` | `/var/run/docker.sock`; lifecycle hooks |
| `/var/run/docker.sock` | Same path; lifecycle hooks |
## Recovery limits
Only the six listed directories are covered, subject to the exclusions above.
Authentik identity/MFA state and Wardrowbe PostgreSQL/uploads are included;
Wardrowbe Redis is excluded. SSD-backed Wardrowbe model/cache paths are outside
the plan. No supplied coverage for Traefik, Backrest configuration, Docker named
volumes, OS, or Mercury. Authelia is retired; old data retention is unconfirmed.
Full-host recovery requires independently retrievable repository credentials,
configuration and restic key. Key custody alone does not establish recoverability.
No independent Google Drive/rclone credential recovery or Backrest-config
recovery arrangement exists. Recovering backed-up Portainer state is not a
substitute for first being able to retrieve the backup.
`/restore` is on the same RAID array; not an independent backup.
- No supplied coverage for Backrest config, Traefik, OS/host config, Mercury, named volumes or SSD model/cache paths. Wardrowbe Redis explicitly excluded.
- No independent Git/PAT, stack-secret, Authentik admin/MFA or Google Drive/rclone recovery arrangement; restic key alone cannot retrieve backups.
- Proposed, **not implemented**: sanitised host/bootstrap config in private Git; encrypted off-host credential/config archive; independently usable access/decryption instructions.
- Restore/retrieval must work without Jupiter or Portainer; recovering Portainer's DB cannot bootstrap access to an inaccessible backup.
- `/restore` is on the same RAID array, not an independent copy. RPO/RTO and full-host recovery order remain unproven.
+57 -116
View File
@@ -4,137 +4,78 @@ id: home-server.deployment
type: reference
scope: [jupiter, portainer, github, stack-deployment, secrets]
sensitivity: private-infrastructure
last_reviewed: "2026-10-26"
last_reviewed: "2026-10-27"
sources:
- kind: owner-report
reference: "AirVPN local downloads and public authenticated qBittorrent UI live; Ultra.cc/rclone retired; encrypted download-state/secret backups not yet added, 2026-10-26"
reference: "Portainer Git/PAT/manual deployment 2026-10-06; Authentik deployed 2026-10-16; downloads deployed 2026-10-26; ingress runbook completed 2026-10-27"
- kind: repository
repository: jupiter-stacks
revision: working-tree
base_revision: 631e4fdfa4c7122e5eff38e11b070c17ff7a4a2a
paths:
- stacks/downloads/docker-compose.yml
- stacks/downloads/README.md
- stacks/sonarr/docker-compose.yml
- stacks/radarr/docker-compose.yml
- stacks/prowlarr/docker-compose.yml
- stacks/traefik/docker-compose.yml
- kind: repository
repository: jupiter-stacks
revision: 026cfa58b601165342b278401183c5fed5a50db2
paths:
- stacks/homarr/docker-compose.yml
- stacks/authentik/docker-compose.yml
- stacks/authentik/blueprints/jupiter.yaml
- stacks/authentik/.env.example
- kind: owner-report
reference: "Portainer GitHub/main/PAT configuration, environment management and manual pull/redeploy workflow, 2026-10-06"
- kind: owner-report
reference: "No independent credential recovery; operational rules not established, 2026-10-06"
- kind: owner-report
reference: "Authentik fully deployed with all listed application integrations; Authelia retired; independent recovery custody unchanged, 2026-10-16"
- kind: repository
repository: jupiter-stacks
revision: d09df456ad331b8beb9a40615114ea8aae2dced5
paths:
- stacks/authentik/docker-compose.yml
- stacks/authentik/blueprints/jupiter.yaml
- stacks/authentik/.env.example
- stacks/beszel/docker-compose.yml
- stacks/gitea/docker-compose.yml
- stacks/immich/docker-compose.yml
- stacks/paperless-ngx/docker-compose.yml
related: [home-server.reference, home-server.host, home-server.portainer, home-server.authentication, home-server.backups, home-server.operations, home-server.notes, home-server.downloads]
update_triggers: [deployment-workflow-change, repository-change, credential-rotation, rollback-policy-change]
unknowns:
- github-pat-scopes-and-expiration
- post-deployment-health-criteria
revision: 12fb1b96e01fc062fdce4c216f11eb5d7ce716ed
paths: [stacks/traefik/docker-compose.yml, stacks/authentik/docker-compose.yml, stacks/downloads/docker-compose.yml, stacks/notes/docker-compose.yml]
related: [home-server.portainer, home-server.authentication, home-server.downloads, home-server.notes, home-server.networking, home-server.backups]
update_triggers: [deployment-workflow-change, repository-change, credential-rotation]
unknowns: [github-pat-scopes-and-expiration, general-health-gates-and-approval-policy]
---
# Deployment
| Setting | Source / workflow |
| --- | --- |
| Stack definitions | <https://github.com/jayo60013/jupiter-stacks> |
| Git ref | `main` |
| Repository authentication | GitHub PAT configured in Portainer |
| Environment values/secrets | Portainer stack environment settings; AirVPN keys use root-restricted host secret files |
| Deployment | Manually pull/redeploy each affected stack in Portainer |
| Portainer itself | [Host shell script](portainer.md), not a managed Compose stack |
```text
Compose change -> push main -> manual Portainer pull/redeploy -> running stack
```mermaid
---
config:
themeCSS: |
text,.nodeLabel { fill:var(--infra-text);color:var(--infra-text); }
.flowchart-link { stroke:var(--infra-line); }
marker path { fill:var(--infra-line);stroke:var(--infra-line); }
.infraClient rect { fill:var(--infra-client-fill);stroke:var(--infra-client-stroke);rx:8px;ry:8px; }
.infraEdge rect { fill:var(--infra-edge-fill);stroke:var(--infra-edge-stroke);rx:8px;ry:8px; }
.infraService rect { fill:var(--infra-service-fill);stroke:var(--infra-service-stroke);rx:8px;ry:8px; }
---
flowchart LR
accTitle: Manual Git-backed stack deployment
accDescr: Pushing configuration does not deploy it; Portainer must pull and redeploy before runtime checks.
Git["Push main<br/>jupiter-stacks"]:::infraClient
Portainer["Portainer<br/>manual pull / redeploy"]:::infraEdge
Runtime["Running containers<br/>check routes + health"]:::infraService
Git --> Portainer --> Runtime
```
A Git commit is not deployment proof. Installed host Compose version is not
proof of Portainer's Compose implementation.
- Source: <https://github.com/jayo60013/jupiter-stacks>, branch `main`; GitHub PAT in Portainer.
- Stack values/secrets live in Portainer; AirVPN keys are root-restricted host files.
- Git updates do not populate bind-mounted files or application databases.
- Native Compose version does not establish Portainer's Compose implementation.
- Host apt, fstab/RAID, systemd and Mercury nginx are outside this workflow.
Recovering deployment requires Portainer state or independently retrievable
repository credentials and stack environment values. [Portainer backups](backups.md)
exist; credential/state restoration is untested. Do not put secret values in
this site or Git.
## Prerequisites / order
No independent recovery arrangement exists for the PAT or stack secrets.
Rollback, health-check gates, approval rules and UI-versus-Git drift policy
are not established; do not assume permission to deploy, stop or delete services.
| Before | Then | Detail |
| --- | --- | --- |
| Mounted `/data`, Docker, required networks | Workloads | Preserve bind paths, named volumes and secrets |
| Downloads stack creates `downloads` | Arr + Traefik | [Download state/redeployment](downloads.md#state-and-redeployment) |
| Gitea creates `notes_build` | Notes | [Notes deployment](notes.md#deployment) |
| Secure Authentik + apply host blueprint | Public authentication | [Authentication](authentication.md#state-secrets) |
| Independent SSH session | Replace Portainer | [Standalone deployment](portainer.md#deployment) |
## Local downloads deployment / recovery
## Ingress transport deployment and rollback
The owner confirms [qBittorrent/Gluetun through AirVPN](downloads.md) live on
2026-10-26, including both Arr import workflows and the authenticated public UI.
Use the linked stack runbook for reconstruction, not the retired seedbox.
Current routing convention: [Networking](networking.md#listener-and-trust-boundaries).
Owner confirmed cutover complete **2026-10-27**.
Deploy Downloads before its consumers: it creates the external `downloads`
network used by Sonarr, Radarr, Prowlarr and Traefik. Prepare the bind-mounted
files under `/data/qbittorrent/{config,defaults}` and
`/data/gluetun/secrets` first. AirVPN private/preshared keys are root-restricted
host files mounted as Compose secrets, not Portainer environment values;
the tunnel address, country and reserved peer port are stack environment values.
1. Keep SSH and configuration backups; ensure WireGuard owns `10.0.0.2`.
2. Deploy Traefik's HTTPS defaults, restricted WireGuard listener and trusted forwarded headers **before** changing app routes.
3. Redeploy public routers as `https,wireguard` without router TLS labels; retain local HTTPS-only routes and HSTS proxy-scheme handling.
4. Recreate Portainer independently; include external football's router. Recreate Gluetun/qBittorrent together; preserve Authentik names used by [backup hooks](backups.md#consistency-hooks).
5. From Mercury, check each host/path at HTTP `10.0.0.2:8081` with original HTTPS headers. Do not follow fallback redirects; verify callbacks/HSTS, non-Mercury denial and LAN HTTPS.
6. Switch Mercury **last**: install the shared upstream plus all three nginx site files together; `sudo nginx -t && sudo systemctl reload nginx`.
7. Check public login/callbacks, WebSockets, registry and media responses.
Arr client credentials, categories, library roots, remote-path mappings and
hardlink settings are database/UI state, not reproduced by pulling Compose.
The public UI also depends on the persisted qBittorrent Server domains setting,
the host-side Authentik blueprint and Traefik's `downloads` network attachment.
Recreate Gluetun and qBittorrent together when replacing their shared namespace.
- Transport rollback: restore Mercury's previous **sites and upstream together** (`https://jupiter_backend` + `10.0.0.2:443`), syntax-check, reload.
- Leave migrated Jupiter routes in place: HTTPS still works. If reverting Traefik too, restore app TLS labels before removing entrypoint defaults.
- Host staging/backups: `~/jupiter-wireguard-http` on both hosts; Mercury's rollback files under `backup/{conf.d,sites-available}`. Not independent recovery storage.
Full isolation exercises remain outstanding, and the new client state/defaults
and VPN secrets are not yet backed up. See [Backups](backups.md#local-downloads-recovery-gap).
## Reconstruction limits
## Notes deployment
The [Notes reference](notes.md) covers `notes_builder`, `notes_web`, host bind
mounts, the signed Gitea webhook and the shared static volume. Deploy Gitea before
Notes: Gitea owns the `notes_build` network that the Notes stack consumes.
## Authentik configuration surfaces
The owner reports the Authelia-to-Authentik migration complete. Rebuilding it
requires more than pulling Compose:
| Surface | Required state |
| --- | --- |
| Portainer environment | Authentik persistent key, database password, eight provider client secrets; matching application secrets/overrides |
| Host files | Copy repository `stacks/authentik/blueprints/jupiter.yaml` into `/data/authentik/blueprints`; Git checkout alone does not populate this bind mount |
| Authentik database | Administrator, MFA enrolment, group memberships, applied providers/application policies |
| Application databases/UIs | Beszel, Gitea, Immich and Portainer OIDC setup/account linking |
| Application Compose | Dozzle, Homarr, Paperless and Wardrowbe OIDC settings; five proxy-protected routers use `authentik@docker` |
Homarr uses the same `OIDC_HOMARR_SECRET` in both stack environments. Its
[account-linking steps](authentication.md#homarr) include verified email and
duplicate-account cleanup; redeploying Compose alone does not complete linking.
For initial setup, keep `AUTHENTIK_PUBLIC_ROUTING=false` and access the
loopback-published `127.0.0.1:9000` privately, for example through SSH forwarding.
Secure the administrator account and prepare MFA, groups and blueprint policy
before enabling public routing. Enable routing and establish working OIDC
account links before disabling any local application login; Beszel/Paperless
flags default to preserving it. The retired Authelia service is not a
replacement recovery path.
Redeploying the Authentik stack alone does not copy changed host blueprints
or reproduce database-backed application settings. Keep provider/client secrets
matched across both sides. Preserve deployed Authentik container names or update
the [Backrest hooks](backups.md) when recreating containers.
Host apt packages, fstab, RAID assembly, systemd units and Mercury nginx are
outside the Compose deployment workflow.
- Restore Portainer state or independently retrieve Git credentials + stack secrets: [Backups](backups.md#recovery-limits).
- Authentik: copy blueprint into its host bind mount; restore identity/MFA and application-managed OIDC state. Compose alone is insufficient.
- Football is externally managed; its Compose definition is not in `jupiter-stacks`.
- Portainer image pin caveat: [Portainer](portainer.md#deployment).
- General approvals/drift policy and full-host recovery order remain undefined.
+73 -152
View File
@@ -2,185 +2,106 @@
schema_version: 1
id: home-server.downloads
type: reference
scope: [jupiter, airvpn, gluetun, qbittorrent, media-import]
scope: [jupiter, airvpn, gluetun, qbittorrent]
sensitivity: private-infrastructure
last_reviewed: "2026-10-26"
last_reviewed: "2026-10-27"
sources:
- kind: owner-report
reference: "Sonarr, Radarr and Prowlarr now have alerting enabled; destination and enabled events not supplied, 2026-10-26"
- kind: owner-report
reference: "Former Ultra.cc subscription active since 2024-05-25, always GBP 4.50/month until retirement, 2026-10-26"
- kind: owner-report
reference: "AirVPN live with direct home-ISP egress; Sonarr/Radarr imports, hardlinks and seeding/removal confirmed; public UI and both login layers working; Ultra.cc and rclone mount retired; remaining isolation checks and backups outstanding, 2026-10-26"
- kind: owner-report
reference: "Cost saving and simpler architecture; three-year AirVPN plan purchased for GBP 56.06, expires Thursday, 25 October 2029, 2026-10-26"
- kind: owner-report
reference: "Requested AirVPN/local qBittorrent migration with direct home-ISP egress, 2026-10-26"
reference: "AirVPN egress, imports/hardlinks, seeding/removal and authenticated UI live; subscription details and Ultra.cc retirement confirmed 2026-10-26; isolation exercises incomplete"
- kind: repository
repository: jupiter-stacks
revision: working-tree
base_revision: 631e4fdfa4c7122e5eff38e11b070c17ff7a4a2a
paths:
- stacks/downloads/docker-compose.yml
- stacks/downloads/README.md
- stacks/downloads/qBittorrent.conf
- stacks/downloads/resolv.conf
- stacks/downloads/vpn-entrypoint.sh
- stacks/sonarr/docker-compose.yml
- stacks/radarr/docker-compose.yml
- stacks/prowlarr/docker-compose.yml
related: [home-server.seedbox, home-server.networking, home-server.storage, home-server.deployment, home-server.backups, home-server.operations]
update_triggers: [deployment, airvpn-device-change, route-change, storage-change, seedbox-retirement, alerting-change]
unknowns:
- selected-airvpn-server-country-and-forwarded-port
- measured-throughput-and-runtime-isolation
- airvpn-renewal-behaviour
- arr-alert-destinations-and-enabled-events
paths: [stacks/downloads/docker-compose.yml, stacks/downloads/qBittorrent.conf, stacks/downloads/resolv.conf, stacks/downloads/vpn-entrypoint.sh, stacks/sonarr/docker-compose.yml, stacks/radarr/docker-compose.yml]
related: [home-server.networking, home-server.storage, home-server.deployment, home-server.backups, home-server.operations, home-server.seedbox]
update_triggers: [airvpn-device-change, route-change, storage-change, renewal]
unknowns: [airvpn-country-and-reserved-port, measured-throughput, complete-runtime-isolation, renewal-behaviour]
---
# Local downloads / AirVPN
**Live by owner report on 2026-10-26.** Local qBittorrent behind Gluetun/AirVPN
has replaced the [Ultra.cc seedbox](seedbox.md). Downloads use the home ISP
directly. Sonarr and Radarr imports, hardlinks, continued seeding and intended
removal behaviour are confirmed. Ultra.cc is cancelled/expired, retained files
have been handled, and the old rclone-seedbox service and FUSE mount are removed.
The change saves money and simplifies the architecture: downloads and libraries
now share local storage, removing the remote seedbox, SFTP/FUSE mount and
cross-filesystem import dependency. This is owner confirmation, not independent
runtime inspection; the remaining isolation and recovery gaps are listed below.
## Subscription / savings
| Item | Cost / term |
| --- | --- |
| Provider | [AirVPN](https://airvpn.org/); three-year plan purchased |
| Upfront payment | GBP 56.06 for three years |
| Expiry | **Thursday, 25 October 2029** |
| Effective subscription cost | Approximately GBP 1.56/month or GBP 18.69/year, amortised over 36 months |
| Former Ultra.cc subscription | Active since 2024-05-25 until retirement; always GBP 4.50/month (GBP 54/year), with no price changes |
| Subscription saving | Approximately GBP 2.94/month or GBP 35.31/year; GBP 105.94 over three years at the former rate |
Savings exclude incremental electricity and any future renewal price.
Automatic renewal behaviour has not been recorded.
The authoritative deployment and rollback runbook is
[`jupiter-stacks/stacks/downloads/README.md`](https://github.com/jayo60013/jupiter-stacks/blob/main/stacks/downloads/README.md).
It includes the required host files, AirVPN setup, routing acceptance gates,
failure exercises, application UI changes and seedbox retirement.
- **Live, owner-confirmed 2026-10-26:** home-ISP egress, Sonarr/Radarr hardlink imports, seeding/removal and both Web UI login layers.
- Replaces [Ultra.cc](seedbox.md); local imports remove the SFTP/FUSE dependency.
- Working downloads do **not** prove fail-closed behaviour; outstanding exercises below.
## Current topology
```text
qBittorrent
-> shared Gluetun network namespace / firewall
-> AirVPN WireGuard packets over Jupiter's home interface/router
-> AirVPN public exit
-> peers
Sonarr / Radarr / optional Prowlarr direct downloads
-> dedicated downloads Docker network
-> http://gluetun:8080
```mermaid
---
config:
themeCSS: |
text,.nodeLabel,.cluster-label { fill:var(--infra-text);color:var(--infra-text); }
.edgeLabel,.edgeLabel p { background:var(--infra-zone);color:var(--infra-text); }
.flowchart-link { stroke:var(--infra-line); }
marker path { fill:var(--infra-line);stroke:var(--infra-line); }
.cluster rect { fill:var(--infra-zone);stroke:var(--infra-border);stroke-dasharray:5 5;rx:12px;ry:12px; }
.cluster span,.cluster text { color:var(--infra-text)!important;fill:var(--infra-text)!important; }
.infraClient rect { fill:var(--infra-client-fill);stroke:var(--infra-client-stroke);rx:8px;ry:8px; }
.infraService rect { fill:var(--infra-service-fill);stroke:var(--infra-service-stroke);rx:8px;ry:8px; }
.infraEdge rect { fill:var(--infra-edge-fill);stroke:var(--infra-edge-stroke);rx:8px;ry:8px; }
.infraCloud rect { fill:var(--infra-cloud-fill);stroke:var(--infra-cloud-stroke);rx:8px;ry:8px; }
---
flowchart LR
accTitle: Download API and VPN-only torrent egress
accDescr: Arr uses the internal authenticated API. qBittorrent shares Gluetun's namespace and reaches peers through AirVPN, not Mercury.
Arr["Sonarr / Radarr / Prowlarr"]:::infraClient
subgraph Namespace["SHARED NETWORK NAMESPACE"]
Qbit["qBittorrent<br/>torrent interface wg0"]:::infraService
VPN["Gluetun<br/>firewall + WireGuard"]:::infraEdge
end
ISP["Home ISP"]:::infraEdge
Exit["AirVPN exit"]:::infraCloud
Peers["Torrent peers"]:::infraClient
Arr -->|"HTTP gluetun:8080 / native login"| Qbit
Qbit --> VPN -->|"Encrypted VPN"| ISP --> Exit <--> Peers
```
The existing Jupiter-to-Mercury host WireGuard tunnel remains separate and
unchanged. Direct home-ISP egress, not Mercury, is confirmed by the owner.
Recheck host routing and packet-marking rules after network or endpoint changes.
## Repository configuration
| Surface | Configuration |
| --- | --- |
| Downloads stack | Pinned Gluetun and LinuxServer qBittorrent images in one Compose project |
| Egress isolation | qBittorrent uses `network_mode: service:gluetun`; enabled Gluetun firewall; initial torrent interface binding `wg0` |
| Startup | Compose health dependency plus a per-start entrypoint guard before qBittorrent's original `/init`, including Docker automatic restarts |
| Namespace | `network_mode: service:gluetun`; enabled firewall; initial torrent binding `wg0` |
| Startup | Compose health dependency + per-start guard before `/init`, including Docker restarts |
| DNS / IPv6 | qBittorrent resolver pinned to Gluetun localhost; namespace IPv6 disabled |
| Administration | HTTPS `qbittorrent.umbra.mom` via Traefik/Authentik direct user bindings + MFA and qBittorrent login; loopback/SSH fallback retained |
| Client API | `gluetun:8080` on the named `downloads` bridge; application authentication retained |
| Peer ingress | AirVPN-reserved TCP/UDP port allowed through the VPN firewall; no host/router publication |
| Secrets | Root-restricted files under `/data/gluetun/secrets`, mounted only into Gluetun |
Prowlarr and FlareSolverr indexer traffic is not moved behind AirVPN. Sonarr,
Radarr and Prowlarr retain their ordinary networks and internet routes.
| Network | Named `downloads` bridge, non-internal, IPv6 disabled; Gluetun, Traefik and Arr attach |
| Peer ingress | AirVPN-reserved TCP/UDP port; allowed by VPN firewall, **not** published on host/router |
| Indexers | Prowlarr/FlareSolverr and Arr ordinary internet traffic stay outside AirVPN |
## Public monitoring
Public access and both authentication layers are working by owner report.
`https://qbittorrent.umbra.mom` uses Mercury's existing wildcard ingress, then
Traefik and Authentik's per-user/MFA policy before qBittorrent's own login.
This is the normal management UI, not a read-only monitoring account.
- Browser: `https://qbittorrent.umbra.mom`; [Mercury ingress](networking.md#application-ingress) -> [Authentik](authentication.md#forward-auth) -> native qBittorrent login.
- Full management UI, **not read-only**. Arr uses the internal API, not the public URL.
- Traefik reaches `gluetun:8080`; router labels belong to Gluetun. qBittorrent gets no extra network/egress path.
- Loopback fallback: `127.0.0.1:8080` via SSH forwarding.
- Persist qBittorrent's **Server domains** before public routing; defaults do not update existing state. Apply the host Authentik blueprint/callback membership.
- Retain native authentication, CSRF and host checks; public DNS points to Mercury, never the VPN exit.
Traefik joins the existing `downloads` network to reach `gluetun:8080`;
Gluetun and qBittorrent gain no independent network or non-VPN egress path.
The router labels belong to Gluetun, which owns the shared namespace.
The Authentik blueprint adds an application bound to `jay` (with other users
assigned manually), a `forward_single`
provider, and embedded-outpost membership; its callback router includes the
new hostname.
## State and redeployment
For recovery or redeployment, follow the runbook's public-monitoring sequence:
update the persisted qBittorrent Server domains setting, apply the host-side Authentik blueprint and
redeploy Authentik, then Traefik and the Downloads pair. Merely changing the
qBittorrent defaults does not update a running installation. Retain native
authentication and CSRF/host protections; do not expose an unauthenticated API.
DNS must resolve the public hostname to Mercury, not to an AirVPN exit.
| Host path | State |
| --- | --- |
| `/data/qbittorrent/config` | Settings, torrent/resume state |
| `/data/qbittorrent/defaults` | Initial config, startup guard and resolver files |
| `/data/gluetun/secrets` | Root-restricted WireGuard private/preshared keys; Gluetun-only mounts |
Sonarr/Radarr/Prowlarr keep using the internal API. Only browser ingress uses
Mercury; torrent internet traffic still uses AirVPN through the home ISP.
- Prepare bind files/keys before deployment; address, country and reserved port are stack variables.
- Deploy Downloads before its network consumers; replace Gluetun/qBittorrent **together**.
- Arr DB/UI state: native API credentials, categories `sonarr`/`radarr`, roots and remote-path mappings. Restore/reconcile before jobs resume.
- Payload/library paths and mount boundaries: [Storage](storage.md#media-mounts).
- Backup coverage: [Backups](backups.md#local-downloads-recovery-gap); alerts: [Operations](operations.md#observability).
## Local storage and imports
## Subscription / savings
Existing host libraries stay at `/data/jellyfin/movies` and
`/data/jellyfin/tv_shows`. New downloads live under
`/data/jellyfin/torrents/{incomplete,movies,tv}`.
| Item | Value |
| --- | --- |
| AirVPN | GBP **56.06 / three years**, expires **2029-10-25** |
| Amortised cost | GBP 1.56/month / 18.69/year |
| Former Ultra.cc | GBP 4.50/month since 2024-05-25; no price changes |
| Saving | GBP 2.94/month / 35.31/year / 105.94 over three years |
Sonarr/Radarr receive one `/data/jellyfin:/media` mount so imports can hardlink
within one filesystem. The replacement library roots are `/media/tv_shows`
and `/media/movies`, rather than `/tv` and `/movies`, without moving
existing host files. Jellyfin's existing `/data/jellyfin:/media` mount stays
unchanged; download folders must not be included in its library roots.
- Excludes incremental electricity and renewal changes; auto-renewal unknown.
qBittorrent receives only `/data/jellyfin/torrents:/media/torrents`, not the
organised media libraries. All three applications see reported download paths
consistently. Repository/runbook categories are `sonarr` and `radarr`, with their
corresponding download folders. Both applications' hardlink imports and intended
seeding/removal behaviour are confirmed by the owner.
## Remaining validation
Deploy Downloads before the consumer stacks: it creates their external
`downloads` network. Updating Compose does not change download-client settings,
categories, roots or remote-path mappings stored in application databases.
During recovery, restore or reconcile that state before resuming jobs.
The runbook's old-queue migration steps are historical now that Ultra.cc is retired.
## Alerting
**Sonarr, Radarr and Prowlarr have alerting enabled**, by owner report on
2026-10-26. Connections are managed in each application's **Settings -> Connect**;
the destinations and selected events have not been recorded.
This does not establish coverage for empty indexer searches, releases rejected
by quality/profile rules, or stalled torrents. A successful search with no
acceptable results is not normally a health failure. Missing-media or
stalled-download alerts need separate monitoring if required.
Existing host/container and Discord notification coverage is documented in
[Operations](operations.md#observability).
## Remaining validation / recovery gaps
- Record the non-secret server country and reserved peer port if needed for
operations; never put device keys in these notes.
- Direct home-ISP egress and normal downloads are confirmed; working inbound
peer connectivity, DNS confinement and IPv6 denial are not yet fully recorded.
- Exercise tunnel failure, unavailable VPN at startup, Gluetun replacement and
host reboot; no unprotected internet fallback is acceptable.
- Add qBittorrent state/defaults and VPN secrets to the actual encrypted backup
policy; the owner confirms they are not yet covered. The documented
six-directory plan does not cover them.
Seedbox retirement and the local import workflow are complete by owner report;
the former seedbox is no longer an available rollback assumption.
The HDD array remains RAID 0 without redundancy. Hardlinks save space but are
not independent copies or backups.
- Record non-secret server country, tunnel address and reserved port.
- Prove inbound peer connectivity, DNS confinement and IPv6 denial.
- Exercise VPN failure, unavailable VPN at startup, Gluetun replacement and host reboot; **no unprotected fallback**.
- The retired seedbox is not a rollback target.
+38 -79
View File
@@ -4,117 +4,76 @@ id: home-server.host
type: reference
scope: [jupiter, hardware, operating-system, docker]
sensitivity: private-infrastructure
last_reviewed: "2026-10-26"
last_reviewed: "2026-10-27"
sources:
- kind: owner-report
reference: "AirVPN local downloads live; Ultra.cc retired and rclone-seedbox service/FUSE mount removed, 2026-10-26"
- kind: owner-report
reference: "Hardware, OS, bare-metal Docker and data-root confirmation, 2026-10-06"
- kind: owner-report
reference: "hostnamectl, timedatectl, user/group records, Docker info/version and systemd units, 2026-10-06"
- kind: owner-report
reference: "Owner does not recognize backup-puller account; no independent host-config archive, 2026-10-06"
- kind: owner-report
reference: "Added local logging driver with max-size 10m and max-file 3 to /etc/docker/daemon.json, 2026-10-20"
related: [home-server.reference, home-server.storage, home-server.networking, home-server.deployment, home-server.operations]
update_triggers: [hardware-change, os-upgrade, kernel-upgrade, docker-reconfiguration, account-change, unit-change]
unknowns:
- exact-gpu-model
- jupiter-firewall-policy
- docker-package-installation-source
- backup-puller-account-purpose
- docker-logging-restart-and-container-recreation
reference: "Hardware/OS/runtime/accounts/systemd snapshot 2026-10-06; daemon logging change 2026-10-20; seedbox unit removed 2026-10-26"
- kind: live-inspection
reference: "GPU PCI ID 1002:68f9 identified as AMD Cedar, 2026-10-27"
related: [home-server.storage, home-server.networking, home-server.operations]
update_triggers: [hardware-change, os-upgrade, kernel-upgrade, docker-reconfiguration, account-change]
unknowns: [exact-gpu-board-model, docker-package-source, backup-puller-purpose, docker-logging-activation]
---
# Host
# Jupiter host
## Hardware / OS
| Component | Configuration |
| --- | --- |
| Host | Jupiter; bare metal |
| OS | Debian GNU/Linux 13 (trixie); x86_64 |
| Platform | Bare metal; Debian 13 (trixie), x86_64 |
| Kernel | `6.12.94+deb13-amd64` |
| CPU | AMD Ryzen 5 5600X; 6 cores / 12 threads; reported 4.65 GHz |
| CPU | Ryzen 5 5600X; 6 cores / 12 threads; reported 4.65 GHz |
| RAM | 1 x 16 GB Corsair Vengeance LPX |
| GPU | AMD Radeon HD 5000/6000/7350/8350 Series; discrete; exact model unresolved |
| Motherboard | Gigabyte B550M DS3H |
| Firmware | `FB`; dated `2023-06-08` |
| GPU | AMD Cedar `1002:68f9`; exact board unknown |
| Board / firmware | Gigabyte B550M DS3H; `FB`, `2023-06-08` |
| PSU | Corsair CV550 |
| SSD | Samsung 860 EVO; 1 TB; OS and SSD-backed caches |
| HDD | 2 x Seagate BarraCuda; 2 TB each; RAID 0 |
| Docker | Native host engine; data root `/var/lib/docker` |
| LAN | `192.168.1.56` |
| WireGuard | `10.0.0.2/24`; Mercury peer/server `10.0.0.1/24` |
| Time | `Europe/London`; RTC UTC; NTP active; clock synchronized at review |
| Time | `Europe/London`; RTC UTC; NTP active/synchronised at inspection |
| Disks / mounts | [Storage](storage.md); network addresses: [Networking](networking.md) |
## Docker runtime
| Setting | Observed value |
2026-10-06 supplied snapshot, not current-version telemetry.
| Setting | Value |
| --- | --- |
| Engine / API | Docker CE `29.8.2` / `1.56` |
| Compose CLI plugin | `v5.6.0`; not proof of Portainer's Compose implementation |
| Engine / API / Compose | CE `29.8.2` / `1.56` / plugin `v5.6.0` |
| containerd / runc | `v2.3.6` / `1.5.1` |
| Storage | `overlayfs`; containerd snapshotter |
| Cgroups | `systemd`; v2 |
| Logging | `local` configured in `/etc/docker/daemon.json`; `max-size: 10m`, `max-file: 3`; activation not confirmed (previously observed default: `json-file`) |
| Firewall backend | `iptables` |
| Runtime security | AppArmor, seccomp, cgroup namespaces |
| Root / storage | `/var/lib/docker`; `overlayfs`, containerd snapshotter |
| Cgroups / firewall | systemd v2 / iptables |
| Security | AppArmor, seccomp, cgroup namespaces |
| Swarm / live restore | Inactive / disabled |
Docker/containerd services and `docker.socket` are enabled.
Docker starts `/usr/bin/dockerd -H fd:// --containerd=/run/containerd/containerd.sock`.
Supplied Docker unit has no explicit `/data` mount dependency.
The former `rclone-seedbox.service` and FUSE mount are removed by owner report
on 2026-10-26; Docker workloads no longer depend on that remote import mount.
- Docker/containerd/socket enabled; dockerd args: `-H fd:// --containerd=/run/containerd/containerd.sock`.
- Supplied Docker unit has no explicit `/data` mount dependency.
- `wg-quick@wg0` active; separate from Gluetun/AirVPN. WireGuard/rclone apt-managed; retired seedbox unit removed.
- No host nginx; `unattended-upgrades` enabled, policy unknown.
### Docker log rotation
On 2026-10-20, the owner reported adding the following logging settings to
`/etc/docker/daemon.json`:
Owner-reported `/etc/docker/daemon.json`, **2026-10-20**:
```json
{
"log-driver": "local",
"log-opts": {
"max-size": "10m",
"max-file": "3"
}
"log-opts": {"max-size": "10m", "max-file": "3"}
}
```
This configures automatic rotation with roughly 30 MB of uncompressed log
retention per container; the `local` driver also compresses rotated logs.
It covers container stdout/stderr, not application log files written inside
containers or mounted volumes.
A Docker daemon restart and recreation of existing containers are required to
apply this default to them; neither is confirmed by the owner report. Restarting
Docker can interrupt workloads because live restore was disabled in the supplied
snapshot. Explicit Compose/container logging settings override the daemon default.
Inspect a container's effective configuration with:
- Roughly 30 MB uncompressed/container; rotated files compressed. Covers stdout/stderr, not application files.
- Activation unconfirmed: restart daemon **and recreate existing containers**; explicit container logging overrides the default.
- Daemon restart interrupts workloads with live restore disabled.
```sh
docker inspect --format '{{json .HostConfig.LogConfig}}' CONTAINER
```
## Host administration
## Accounts
| Account / group | Identity |
| Identity | Configuration |
| --- | --- |
| `jay` | UID/GID `1000:1000`; `/home/jay`; bash; member of `sudo`, `docker`, `video`, `backups` |
| `backup-puller` | UID/GID `1001:1001`; `/home/backup-puller`; bash; member of `backups`; origin/purpose not recognized by owner |
| `backups` | Shared group; GID `1002` |
| `jay` | UID/GID `1000:1000`; `/home/jay`; bash; `sudo,docker,video,backups` |
| `backup-puller` | UID/GID `1001:1001`; `/home/backup-puller`; bash; group `backups`; purpose unrecognised |
| `backups` | GID `1002` |
WireGuard and rclone were apt-managed in the supplied snapshot;
removing the seedbox service does not establish that the rclone package was
uninstalled. Backrest's separate rclone configuration is not the retired mount.
`wg-quick@wg0.service` is active in the supplied snapshot and remains separate
from the containerised Gluetun/AirVPN tunnel for [local downloads](downloads.md).
No host-level nginx on Jupiter; nginx still exists in application
containers. `unattended-upgrades.service` is enabled; effective update policy
not supplied.
No independent host-configuration archive is maintained. `backup-puller` must
not be assumed to perform backups; identify SSH/cron/service dependencies before
removing or repurposing it.
Hardware/versions are a dated snapshot, not live telemetry.
- Identify `backup-puller` SSH/cron/service dependencies before removal; its name is not backup evidence.
-229
View File
@@ -1,229 +0,0 @@
---
schema_version: 1
id: home-server.ideas-next-steps
type: plan
scope: [jupiter, mercury, private-access, recovery, downloads, power, services]
sensitivity: private-infrastructure
last_reviewed: "2026-10-26"
sources:
- kind: owner-report
reference: "AirVPN downloads live; GBP 56.06 three-year plan ending 2029-10-25; Ultra.cc/rclone retired; imports/hardlinks/seeding/removal and public UI confirmed; full isolation checks and backups outstanding, 2026-10-26"
- kind: owner-report
reference: "Access, device-loss recovery, hardware constraints, broadband and VPN preferences interview, 2026-10-22"
- kind: documentation
reference: "Planning discussion based on jupiter-stacks and home-server reference notes, 2026-10-22; proposals only, no live inspection or deployment"
related:
- home-server.reference
- home-server.networking
- home-server.authentication
- home-server.backups
- home-server.operations
- home-server.seedbox
- home-server.downloads
- home-server.storage
- home-server.host
update_triggers: [implementation, priority-change, access-requirements-change, pricing-change, power-measurement]
unknowns:
- live-routing-firewall-and-wireguard-configuration
- work-laptop-local-dns-and-lan-access-policy
- measured-wall-power-and-electricity-tariff
- airvpn-renewal-terms-and-complete-isolation-results
- immich-original-photo-backup-coverage
---
# Ideas / next steps
Planning backlog from the 2026-10-22 discussion, updated with the owner's
2026-10-26 AirVPN cutover confirmation. Unchecked proposals remain
**recommendations, not implemented changes**; the completed downloads work
below is owner-reported deployed state.
## Requirements confirmed with the owner
| Area | Requirement / preference |
| --- | --- |
| Private administration | Portainer, Beszel, Dozzle, Backrest, Prowlarr, Radarr and Sonarr should stop being publicly reachable |
| Public applications | Keep normal user-facing applications public, including Vaultwarden |
| Personal devices | Owner-only admin access; VPN clients or certificates can be installed on personal phone/laptop |
| Work laptop | Browser-only admin access needed on home Wi-Fi; cannot install VPN clients or certificates |
| Device loss | Vaultwarden access remains essential; admin tools can wait until returning home |
| Vaultwarden recovery | Owner knows the master password; TOTP is backed up/restorable through Microsoft, with independently recoverable Microsoft access; owner report, not an observed recovery exercise |
| Private-access tooling | Prefer minimal infrastructure and no vendor-hosted coordination service; existing Cloudflare DNS/SSH use is acceptable |
| Hardware | Software/BIOS tuning only; no hardware purchases or removals |
| Broadband | Reported 500 Mbps download and upload; public trackers only |
| Download VPN | AirVPN selected and live; three-year plan purchased for GBP 56.06, expires Thursday, 25 October 2029 |
| New services | Interested in reliability, everyday tools, media, development and privacy |
## Priority order
1. Prepare independent recovery access, then introduce private admin ingress.
2. Make backups recoverable and add failure notifications.
3. Finish download-VPN isolation checks and recovery coverage; seedbox replacement is complete.
4. Measure and tune power use.
5. Add services that solve an actual day-to-day problem.
## Private admin access
### Preferred approach: plain WireGuard through Mercury
Reuse Mercury's public address and existing WireGuard connection to Jupiter.
Give each personal device its own peer/key, narrowly scoped routes and firewall
permissions. No Tailscale account or self-hosted coordination service is needed.
| Client | Intended access path |
| --- | --- |
| Work laptop on home Wi-Fi | LAN directly to private admin ingress; browser only |
| Personal device away from home | WireGuard to Mercury, then routed private access to Jupiter |
| Public internet | Mercury nginx to explicitly public applications only |
Use split DNS: admin hostnames resolve to private addresses on the LAN and
through the VPN. Keep HTTPS certificates using the existing DNS-based challenge;
certificate issuance does not require a public website.
Before choosing exact addresses/listeners, inspect live WireGuard `AllowedIPs`,
routing, firewall rules and Docker port publication. Check that the work laptop's
corporate DNS/proxy policy permits local access; do not bypass employer controls.
### Implementation checklist
- [ ] Prepare recovery access and a rollback route before restricting ingress.
- [ ] Add per-device WireGuard peers and a way to revoke a lost device's key.
- [ ] Configure LAN/VPN DNS for the private admin hostnames.
- [ ] Create private-only Traefik ingress; public requests must not reach admin
backends simply by supplying their hostnames.
- [ ] Replace or constrain Mercury's wildcard forwarding in
`jupiter-stacks/mercury/nginx/umbra.mom`, retaining all intended public apps.
- [ ] Apply the same boundary to Authentik's separate outpost callback routes
for Backrest and the Arr applications.
- [ ] Remove or restrict Portainer's direct `9123:9000` publication in
`stacks/portainer/portainer_start.sh`.
- [ ] Retain application authentication and Authentik MFA behind the private
boundary; VPN connectivity must not imply application administrator rights.
- [ ] Confirm private apps and authentication callbacks work from LAN and VPN,
but cannot be reached through public ingress or alternative published ports.
- [ ] Confirm public applications, especially Vaultwarden, remain accessible.
Removing public DNS records or adding a login page is not sufficient isolation.
The network/proxy boundary must enforce the restriction.
### Alternatives
| Option | Assessment |
| --- | --- |
| Plain WireGuard | Preferred: reuse existing infrastructure, few devices, no hosted control service |
| Tailscale | Convenient, but adds the vendor dependency the owner prefers to avoid |
| Headscale | Self-hosted coordination, but more infrastructure to maintain than currently needed |
| mTLS | Device-certificate authentication; adds certificate distribution/recovery and cannot be used directly on the work laptop |
mTLS rejects clients without a certificate but leaves the HTTPS endpoint
internet-facing. Mercury terminates public TLS, so browser client-certificate
enforcement would need to happen there, not merely between Mercury and Jupiter.
## Recovery and reliability
Keep Vaultwarden public with application login and MFA. Do not make access
depend on the new admin VPN or Authentik.
- [ ] Maintain an encrypted recovery pack outside Jupiter: Mercury provider
console/SSH access, WireGuard re-enrolment instructions, Authentik recovery,
retained local admin credentials, backup credentials, restic key, stack
secrets and essential host configuration.
- [ ] Ensure retrieving/decrypting that pack does not depend on Jupiter,
Portainer or secrets available only inside Vaultwarden.
- [ ] Retain the existing Cloudflare SSH route during migration, with its
effective Access policy reviewed. It still depends on Jupiter's connector
and networking; it is not an out-of-band console.
- [ ] Restore an important service into an isolated location and record the
dependencies, time taken and acceptable data loss.
- [ ] Review backup coverage for Backrest configuration, host configuration,
Docker named volumes and Mercury.
- [ ] Establish whether Immich contains irreplaceable originals and where those
originals and the associated database are backed up.
- [ ] Make failed backup restart hooks visible; a successful snapshot must not
hide stopped Authentik, Portainer or Wardrowbe services.
- [ ] Add service, disk-health, capacity and backup-failure notifications using
Beszel and the existing Mercury uptime service where suitable.
- [ ] Add an external missed-backup heartbeat so Jupiter going offline cannot
prevent notification of its own failure.
Private connectivity does not fix Authentik downtime. The documented backup plan
stops Authentik and ignores some restart failures; independent SSH/local recovery
remains necessary. See [Backups](backups.md) and [Operations](operations.md).
## Local downloads instead of the seedbox
**Completed by owner report on 2026-10-26:**
[local qBittorrent/Gluetun downloads through AirVPN](downloads.md) are live.
This is both a **cost saving** and an **architecture simplification**: local
storage and hardlink imports replace the remote seedbox and SFTP/FUSE dependency.
The purchased **three-year AirVPN plan cost GBP 56.06** and expires
**Thursday, 25 October 2029**. Amortised over 36 months, that is approximately
GBP 1.56/month versus Ultra.cc's former GBP 4.50/month: a subscription saving of
**GBP 35.31/year**, or **GBP 105.94 over three years**, before extra electricity.
The former GBP 2.50/month VPN estimate is superseded by this actual purchase.
### Completed work and remaining follow-up
- [x] Deploy local qBittorrent behind Gluetun/AirVPN.
- [x] Confirm encrypted VPN egress uses the home ISP directly, not Mercury.
- [x] Confirm Sonarr and Radarr imports, hardlinks and intended seeding/removal.
- [x] Confirm public Web UI through Mercury/Traefik/Authentik plus native
qBittorrent login; this is an intentional change from the private-only proposal.
- [x] Retire Ultra.cc, handle retained files and remove the rclone-seedbox
service and FUSE mount.
- [ ] Complete peer-ingress, DNS/IPv6 leak and VPN-failure/startup/reboot checks.
- [ ] Add qBittorrent state/defaults and AirVPN secrets to encrypted off-host backups.
A hardlink gives the same file data two names without storing a second full
copy. Hardlinks are confirmed for both import workflows; they are not backups.
The separate private-admin-access proposal above remains outstanding.
## Electricity and compute efficiency
Reported average CPU usage is approximately 1%, but CPU utilisation does not
measure wall-socket power. No power measurement or actual electricity tariff
was supplied.
Illustrative costs at **25p/kWh**, using 365 days/year divided by 12:
| Continuous average draw | Approximate monthly cost |
| --- | --- |
| 40 W | GBP 7.30 |
| 60 W | GBP 10.95 |
| 80 W | GBP 14.60 |
At that illustrative tariff, saving 10 W continuously saves about GBP 18/year.
An extra average draw of about 16 W would consume the GBP 35.31/year subscription saving.
Count the incremental download workload, not the server's existing baseline.
- [ ] Measure wall power and obtain the actual unit rate when practical.
- [ ] Inspect CPU idle states and frequency-scaling policy before BIOS tuning;
low CPU limits do not automatically save energy.
- [ ] Inspect RAM pressure and swapping during Immich processing and Wardrowbe
Ollama jobs; the host has 16 GB RAM.
- [ ] Schedule heavy jobs where useful and avoid unnecessary repeated scans.
- [ ] Review effective container log rotation; daemon defaults require activation
and recreation, and explicit Compose settings override them.
- [ ] Avoid aggressive HDD spindown while `/data` hosts active databases.
- [ ] Avoid consolidating unrelated application databases solely to reduce
container count; shared failures/upgrades may outweigh uncertain savings.
No hardware changes are proposed. Leave spare CPU capacity idle unless a useful
workload justifies the added power and maintenance.
## Optional service ideas
These are candidates, not a deployment list.
| Area | Idea | Constraint / reason |
| --- | --- | --- |
| Reliability | Notifications and external backup heartbeats | Start with existing monitoring rather than a second metrics stack |
| Privacy | AdGuard Home or Pi-hole | Choose one; household DNS availability becomes an operational dependency |
| Files | Syncthing for selected folders | Sync is not backup; check client support on the owner's devices |
| Bookmarks / reading | Linkding; FreshRSS | Small tools with a clear everyday purpose |
| Development | Renovate automation for existing Gitea workflows | Propose updates; avoid unattended production upgrades |
| Media | Profilarr for supported Arr profiles/custom formats | Add only if manual profile maintenance is a real problem |
Defer Kubernetes, a large metrics stack, another identity provider or a general
cloud suite unless a concrete need emerges.
+27 -49
View File
@@ -2,64 +2,42 @@
schema_version: 1
id: home-server.reference
type: reference
scope: [jupiter, mercury, downloads, shared-infrastructure]
scope: [jupiter, mercury, shared-infrastructure]
sensitivity: private-infrastructure
last_reviewed: "2026-10-26"
last_reviewed: "2026-10-27"
sources:
- kind: owner-report
reference: "AirVPN local downloads live, Ultra.cc/rclone retired; GBP 56.06 three-year plan ending 2029-10-25 for savings and simpler architecture, 2026-10-26"
- kind: owner-report
reference: "Added ideas/next-steps planning page from owner discussion, 2026-10-22"
- kind: owner-report
reference: "Owner interview, 2026-10-06"
- kind: owner-report
reference: "Authentik migration, access-group motivation and six-directory backup plan, 2026-10-16"
related:
- home-server.host
- home-server.mercury
- home-server.seedbox
- home-server.downloads
- home-server.portainer
- home-server.deployment
- home-server.notes
- home-server.operations
- home-server.networking
- home-server.storage
- home-server.authentication
- home-server.backups
- home-server.ideas-next-steps
update_triggers: [topology-change, source-change, infrastructure-change]
reference: "Host inventory 2026-10-06; Authentik 2026-10-16; downloads 2026-10-26; ingress cutover 2026-10-27"
related: [home-server.host, home-server.mercury, home-server.networking, home-server.storage, home-server.authentication, home-server.downloads, home-server.deployment, home-server.portainer, home-server.notes, home-server.backups, home-server.operations, home-server.seedbox]
update_triggers: [topology-change, source-change]
unknowns: []
---
# Home server
| Reference | Scope |
Each topic has one reference page; link to it rather than copying its details.
| Topic | Canonical reference |
| --- | --- |
| [Host](host.md) | Jupiter hardware; OS; Docker |
| [Mercury VPS](mercury.md) | Hosting; ingress; WireGuard; firewall |
| [Local downloads / AirVPN](downloads.md) | Live qBittorrent/Gluetun; local hardlink imports; three-year plan GBP 56.06, expires Thursday, 25 October 2029 |
| [Seedbox (retired)](seedbox.md) | Historical Ultra.cc setup; subscription and rclone/FUSE mount retired |
| [Portainer](portainer.md) | Script-managed container; state; backups |
| [Deployment](deployment.md) | GitHub; Portainer; environment values |
| [Notes](notes.md) | Webhook builder; static site; networks; deployment order |
| [Operations / decisions](operations.md) | Tradeoffs; monitoring; recovery gaps |
| [Networking](networking.md) | Mercury; WireGuard; DNS; ingress |
| [Storage](storage.md) | SSD; RAID 0; mounts; volumes |
| [Authentication](authentication.md) | Authentik; OIDC; per-user service access; forward-auth; secrets |
| [Backups](backups.md) | Backrest; coverage; retention; recovery |
| [Ideas / next steps](ideas-next-steps.md) | Proposed private access, recovery, savings and optional services; not deployed state |
| Hardware, OS, Docker and accounts | [Jupiter host](host.md) |
| VPS, host firewall, Fail2ban and Kuma runtime | [Mercury](mercury.md) |
| Traffic paths, DNS, listeners and proxy trust | [Networking](networking.md) |
| Disks, mounts, volumes and media paths | [Storage](storage.md) |
| OIDC, forward-auth, access policy and account linking | [Authentication](authentication.md) |
| VPN isolation, download configuration and subscription | [Local downloads / AirVPN](downloads.md) |
| Git workflow, prerequisites and ingress rollout | [Deployment](deployment.md) |
| Standalone container and persisted management state | [Portainer](portainer.md) |
| Documentation builder and publication | [Notes](notes.md) |
| Coverage, consistency hooks and recovery gaps | [Backups](backups.md) |
| Alerts, troubleshooting and outstanding work | [Operations](operations.md) |
| Retired remote-download service | [Seedbox archive](seedbox.md) |
## Evidence / agent contract
## Evidence conventions
- `owner-report`: supplied host state/configuration; not independently inspected.
- `owner-report`: supplied state; not independent inspection.
- `repository`: intended configuration at `revision`; not deployment proof.
- `documentation`: upstream behaviour; not proof of deployed settings.
- `revision: working-tree`: uncommitted source; `base_revision`: base commit.
- `unknowns`: unresolved facts; never infer defaults as deployed state.
- `related`: stable document IDs; `update_triggers`: re-review conditions.
- On change: update affected prose, Mermaid, sources, unknowns, and `last_reviewed`.
- No credentials. Preserve evidence distinctions; no live status implied.
Mermaid renders from versioned text; diagrams are manually maintained, not
automatically discovered.
- `live-inspection`: dated checks, limited to the recorded scope.
- `documentation`: upstream behaviour, not deployed settings.
- `working-tree` / `base_revision`: uncommitted configuration and its base.
- `unknowns`: unresolved facts; diagrams are manually maintained snapshots.
- Update sources and `last_reviewed` with changes; never store credentials here.
+51 -230
View File
@@ -2,253 +2,74 @@
schema_version: 1
id: home-server.mercury
type: reference
scope: [mercury, vps, hosting, nginx, wireguard, firewall]
scope: [mercury, vps, firewall, fail2ban, uptime-kuma]
sensitivity: private-infrastructure
last_reviewed: "2026-10-26"
last_reviewed: "2026-10-27"
sources:
- kind: owner-report
reference: "Fasthosts plan, fastfetch, nginx sites, iptables and fail2ban output, 2026-10-06"
- kind: owner-report
reference: "nginx/WireGuard apt management and public IPv4/CGNAT rationale, 2026-10-06"
- kind: owner-report
reference: "Uptime service confirmed as Uptime Kuma; notifications received in Discord server Jupiter, channel #uptimekuma-alerts, 2026-10-23"
- kind: owner-report
reference: "Mercury uptime-kuma.sh supplied with bind mount, resource limits and log rotation, 2026-10-26"
- kind: owner-report
reference: "Mercury Uptime Kuma only checks WireGuard tunnel health and whether Jupiter is online; alerts Jay on Discord when down, 2026-10-26"
- kind: owner-report
reference: "Mercury monitored in Beszel; CPU above 90% for 10 minutes, memory above 90% for 10 minutes and any failed systemd services alerts configured, 2026-10-26"
- kind: owner-report
reference: "fail2ban-client version and status output: Fail2ban 1.1.0 with three active jails (nginx-botsearch, nginx-http-auth, sshd), 2026-10-26"
- kind: owner-report
reference: "fail2ban-client get output for each jail: maxretry, findtime, bantime and ignoreip, 2026-10-26"
- kind: owner-report
reference: "fail2ban-client logpath, journalmatch and actions output: sshd uses journal match _SYSTEMD_UNIT=ssh.service + _COMM=sshd; nginx jails use error.log/access.log; all three use nftables action, 2026-10-26"
- kind: owner-report
reference: "Non-comment jail overrides supplied from /etc/fail2ban/jail.local and /etc/fail2ban/jail.d/defaults-debian.conf, 2026-10-26"
- kind: owner-report
reference: "Fail2ban logtarget /var/log/fail2ban.log, dbfile /var/lib/fail2ban/fail2ban.sqlite3, dbpurgeage 86400 seconds and systemctl is-enabled output enabled, 2026-10-26"
reference: "Hosting/firewall snapshot 2026-10-06; Fail2ban CLI/config and Kuma runtime 2026-10-26"
- kind: live-inspection
reference: "Hostname mercury and nginx 1.26.3 inspected 2026-10-27; firewall/hosting not revalidated"
- kind: repository
repository: jupiter-stacks
revision: working-tree
paths:
- mercury/stacks/uptime-kuma/docker-compose.yml
- mercury/stacks/uptime-kuma/README.md
related: [home-server.reference, home-server.host, home-server.networking, home-server.operations]
update_triggers: [hosting-renewal, vps-resize, os-upgrade, proxy-change, firewall-change, certificate-change, alerting-change]
unknowns:
- renewal-price-and-renewal-date
- creation-timestamp-timezone
- certificate-renewal-method
- wireguard-allowedips-and-keepalive
- backup-and-restore-policy
- uptime-kuma-exact-probes-and-alert-timing
- fail2ban-ban-escalation-action-details-log-rotation-and-recovery-access
paths: [mercury/stacks/uptime-kuma/docker-compose.yml]
related: [home-server.networking, home-server.operations, home-server.backups]
update_triggers: [hosting-renewal, vps-resize, os-upgrade, firewall-change, kuma-change]
unknowns: [renewal-price-and-date, creation-timestamp-timezone, certificate-renewal-method, mercury-backup-policy, fail2ban-enforcement-details-and-log-retention]
---
# Mercury VPS
## Host / subscription
| Field | Configuration |
| --- | --- |
| Provider | [Fasthosts](https://www.fasthosts.co.uk/); UK data centre |
| Role / OS hostname | Mercury / `my-vps` |
| Plan | `1-1-10`; KVM virtual machine |
| CPU | 1 vCore; reported AMD EPYC-Milan @ 2.00 GHz |
| RAM | 1 GB plan; guest reports 864.37 MiB |
| Disk | 10 GB NVMe SSD plan; guest root ext4, 9.64 GiB |
| OS | Debian GNU/Linux 13 (trixie); x86_64 |
| Kernel | `6.12.85+deb13-amd64` |
| Swap | Disabled |
| Public interface | `ens6`; `185.230.217.66/32` |
| WireGuard | Server `10.0.0.1/24`; Jupiter peer `10.0.0.2/24` |
| Created | `2026-03-01 20:48:51`; timezone unconfirmed |
| Payment | GBP 10.00 prepaid for one year; renewal terms unconfirmed |
Capacities distinguish provider allocation from guest-visible values.
## Ingress
nginx runs directly in the VM, not in Docker; nginx and WireGuard are
apt-managed. Mercury provides static public IPv4 ingress because Jupiter's
home connection uses CGNAT. Residential-IP exposure is avoided by design.
| Hostname | Upstream |
| --- | --- |
| `umbra.mom`, `*.umbra.mom` | `https://10.0.0.2:443` over WireGuard |
| `gitea.umbra.mom` | Same; dedicated registry location `/v2/` |
| `john.gatward.dev`, `samstoreymusic.com` | `https://10.0.0.2:443` |
| `uptime.umbra.mom` | `http://127.0.0.1:3001` |
- HTTP -> HTTPS: `301`; unmatched default server: `418`.
- TLS terminates at nginx; upstream HTTPS terminates again at Jupiter Traefik.
- Upstream certificate verification disabled: `proxy_ssl_verify off`.
- Certificates: `/etc/letsencrypt/live/<domain>/{fullchain.pem,privkey.pem}`;
`umbra.mom` certificate also serves its subdomains.
- Preserves `Host`; sets `X-Real-IP`, `X-Forwarded-For`, `X-Forwarded-Proto`;
forwards WebSocket upgrade headers.
- Gitea `/v2/`: unlimited request body; proxy timeouts `900s`.
## Monitoring / notifications
Mercury is now monitored by the **Beszel hub on Jupiter**, confirmed by the
owner on 2026-10-26. The following Mercury alerts are configured:
| Alert | Trigger |
| --- | --- |
| CPU usage | Above 90% for 10 minutes |
| Memory usage | Above 90% for 10 minutes |
| Failed Services | Any failed systemd service |
These are deployed alert settings, not proposed additions. Beszel's Failed
Services rule detects failed systemd units, not every error-level journal entry.
Alert settings are stored in the Beszel hub UI/database, not the agent Compose
file. Beszel cannot deliver these alerts while Jupiter or its hub is down.
Uptime Kuma runs on Mercury, served at `https://uptime.umbra.mom` through
the local upstream `http://127.0.0.1:3001`. Its current monitoring scope is
**only the health of the WireGuard tunnel and whether Jupiter is online**.
It notifies **Jay on Discord when either is down**, using the **Jupiter**
Discord server's `#uptimekuma-alerts` channel. Coverage confirmed by the owner
on 2026-10-26; notification receipt previously confirmed on 2026-10-23.
It does not currently monitor individual containers or application health.
The supplied `~/uptime-kuma.sh` is now represented by
`mercury/stacks/uptime-kuma/docker-compose.yml` in `jupiter-stacks`.
It retains image `louislam/uptime-kuma:2-slim`, container name `uptime-kuma`,
restart policy `unless-stopped`, loopback port `3001`, and bind mount
`/opt/uptime-kuma:/app/data`. Limits remain 300 MiB RAM and 0.5 CPU, with
10 MiB / 3-file log rotation (Compose explicitly selects `json-file`).
The adjacent README documents a stopped-data backup and migration into the
Mercury Portainer environment. Migration has not been confirmed; the repository
change alone does not replace the running container. nginx remains host-managed.
Exact probe types, targets and alert timing have not been recorded. A failed
check over WireGuard can indicate a tunnel/path failure rather than a powered-off
Jupiter. Mercury monitors independently of Jupiter, but cannot report its own
complete outage through this instance; an external monitor is not yet confirmed.
See [Observability](operations.md#observability) for container-alerting guidance.
## Firewall / fail2ban
- UFW-managed iptables: INPUT/FORWARD `DROP`; OUTPUT `ACCEPT`.
- UFW allows TCP `22,80,443`; UDP `80,443,51820`.
- Docker forwarding has separate rules; INPUT policy alone does not define
container exposure.
- Fail2ban version: `1.1.0`; three active jails: `nginx-botsearch`,
`nginx-http-auth`, `sshd`. Confirmed from owner-supplied CLI output on
2026-10-26.
Read-only commands to check the installed version and active jails:
```bash
sudo fail2ban-client version
sudo fail2ban-client status
```
Effective jail settings from owner-supplied CLI output on 2026-10-26:
| Jail | Retry threshold (`maxretry`) | Detection window (`findtime`) | Ban duration (`bantime`) | IP/network exemptions (`ignoreip`) |
| --- | --- | --- | --- | --- |
| `sshd` | 5 | 600 seconds (10 minutes) | 600 seconds (10 minutes) | None |
| `nginx-http-auth` | 3 | 600 seconds (10 minutes) | 600 seconds (10 minutes) | None |
| `nginx-botsearch` | 2 | 600 seconds (10 minutes) | 3600 seconds (1 hour) | None |
These thresholds count filter-matched failures within the detection window,
not arbitrary requests. No IP/network is explicitly exempt through `ignoreip`
in any of the three jails; other ignore mechanisms have not been inspected.
Ban escalation settings and enforcement action details have not yet been
recorded. Avoid deliberate failed SSH login
tests without a recovery route: the administrator's IP has no `ignoreip`
exemption.
### Fail2ban log sources / enforcement
Owner-supplied runtime output on 2026-10-26 confirms:
| Jail | Monitored source | Configured action |
| --- | --- | --- |
| `sshd` | systemd journal: `_SYSTEMD_UNIT=ssh.service + _COMM=sshd`; no log file monitored | `nftables` |
| `nginx-http-auth` | `/var/log/nginx/error.log`; no journal match set | `nftables` |
| `nginx-botsearch` | `/var/log/nginx/access.log`; no journal match set | `nftables` |
The `+` in the SSH journal match is an OR: entries matching `ssh.service` or
the `sshd` command are selected. All three jails use the `nftables` action;
the action name alone does not establish the exact blocked ports, chains or
IPv4/IPv6 coverage. Those action parameters have not yet been inspected.
### Fail2ban persistent configuration
Owner-supplied non-comment configuration on 2026-10-26:
`/etc/fail2ban/jail.local`:
```ini
[nginx-botsearch]
enabled = true
port = http,https
filter = nginx-botsearch
logpath = /var/log/nginx/access.log
maxretry = 2
findtime = 600
bantime = 3600
[nginx-http-auth]
enabled = true
filter = nginx-http-auth
port = http,https
logpath = /var/log/nginx/error.log
maxretry = 3
```
`/etc/fail2ban/jail.d/defaults-debian.conf`:
```ini
[DEFAULT]
banaction = nftables
banaction_allports = nftables[type=allports]
[sshd]
backend = systemd
journalmatch = _SYSTEMD_UNIT=ssh.service + _COMM=sshd
enabled = true
```
The nginx jails explicitly target `http,https`. Defining
`banaction_allports` does not establish that any active jail uses all-port
banning. The SSH thresholds and nginx HTTP auth timing in the runtime table
are not explicitly set in these snippets; they are inherited from other
configuration/defaults.
These host configuration files are outside the Docker Compose stacks. Their
backup coverage and any direct edits to `jail.conf` or filter/action files
have not been confirmed.
### Fail2ban logging / history
Owner-supplied runtime settings on 2026-10-26:
## Host
| Setting | Value |
| --- | --- |
| Fail2ban log destination (`logtarget`) | `/var/log/fail2ban.log` |
| Persistent database (`dbfile`) | `/var/lib/fail2ban/fail2ban.sqlite3` |
| Database purge age (`dbpurgeage`) | 86400 seconds (24 hours) |
| Start at boot (`systemctl is-enabled fail2ban`) | `enabled` |
| Provider / plan | Fasthosts UK; KVM `1-1-10` |
| CPU / RAM | 1 vCore, AMD EPYC-Milan reported; 1 GB plan / 864.37 MiB guest |
| Disk | 10 GB NVMe plan / 9.64 GiB ext4 root; no swap |
| OS / kernel | Debian 13 x86_64; `6.12.85+deb13-amd64` |
| Hostname | `mercury`; older snapshot: `my-vps` |
| Public interface | `ens6`; `185.230.217.66/32` |
| Created / payment | `2026-03-01 20:48:51`, timezone unknown; GBP 10 prepaid/year |
| Administration | From Jupiter: `ssh mercury`; nginx/WireGuard apt-managed |
| Certificates | `/etc/letsencrypt/live/<domain>/{fullchain.pem,privkey.pem}`; renewal method unknown |
The purge age controls cleanup of old database ban history; it is not the ban
duration and does not define log-file retention. Log rotation and retained
log history have not yet been inspected. The database is not an established
long-term event archive, so a future historical ban-count widget must account
for cleanup and log rotation rather than treating it as complete history.
- Ingress, tunnel addresses and nginx configuration: [Networking](networking.md).
- Alert scope, thresholds and Discord destinations: [Operations](operations.md#observability).
Boot enablement does not by itself prove service health or successful
enforcement. Current bans can be inspected without changing configuration:
## Uptime Kuma runtime
```bash
- Supplied launch script: `~/uptime-kuma.sh`; Compose replacement in `mercury/stacks/uptime-kuma/docker-compose.yml` is **not confirmed deployed**.
- Image `louislam/uptime-kuma:2-slim`; container `uptime-kuma`; restart `unless-stopped`.
- Bind `/opt/uptime-kuma:/app/data`; loopback `127.0.0.1:3001`.
- Limits: **300 MiB / 0.5 CPU**; `json-file` logs **10 MiB x 3**.
- Stop before copying state for migration; preserve the bind directory. nginx stays host-managed.
## Firewall / Fail2ban
- Supplied firewall: UFW/iptables INPUT + FORWARD `DROP`; OUTPUT `ACCEPT`.
- Allows TCP `22,80,443`; UDP `80,443,51820`. Docker forwarding has separate rules.
- Fail2ban **1.1.0**, enabled at boot; effective settings from **2026-10-26**:
| Jail | Source | Retries / window | Ban |
| --- | --- | --- | --- |
| `sshd` | journal: `_SYSTEMD_UNIT=ssh.service + _COMM=sshd` | 5 / 600s | 600s |
| `nginx-http-auth` | `/var/log/nginx/error.log` | 3 / 600s | 600s |
| `nginx-botsearch` | `/var/log/nginx/access.log` | 2 / 600s | 3600s |
- All three: action `nftables`; `ignoreip` empty. SSH journal `+` means **OR**.
- nginx jails target `http,https`. Action names do not prove blocked chains, ports or IPv6 coverage.
- Persistent overrides: `/etc/fail2ban/jail.local` (nginx filters/paths/retries; botsearch timing) and `jail.d/defaults-debian.conf` (SSH systemd backend/match; nftables defaults).
- `banaction_allports=nftables[type=allports]` is defined; active use is unconfirmed. Other timing values are inherited.
- Log: `/var/log/fail2ban.log`; DB: `/var/lib/fail2ban/fail2ban.sqlite3`; `dbpurgeage=86400`.
- DB purge age is **not** ban duration or log retention; escalation, rotation and config backup remain unknown.
- No administrator IP exemption: retain recovery access before testing failed logins.
```sh
sudo fail2ban-client version
sudo fail2ban-client status
sudo fail2ban-client status sshd
sudo fail2ban-client status nginx-http-auth
sudo fail2ban-client status nginx-botsearch
```
Rules and versions are supplied snapshots, not live inspection.
+104 -248
View File
@@ -2,188 +2,124 @@
schema_version: 1
id: home-server.networking
type: reference
scope: [jupiter, mercury, airvpn, downloads, dns, ingress, wireguard, cloudflare]
scope: [jupiter, mercury, dns, ingress, wireguard, cloudflare]
sensitivity: private-infrastructure
last_reviewed: "2026-10-26"
last_reviewed: "2026-10-27"
sources:
- kind: owner-report
reference: "AirVPN local downloads and direct home-ISP egress live; public qBittorrent UI and both authentication layers working; Ultra.cc/rclone mount retired; full failure/leak checks outstanding, 2026-10-26"
reference: "DNS/Cloudflare/network snapshots 2026-10-06; ingress cutover verified 2026-10-27"
- kind: live-inspection
reference: "Traefik/nginx configuration, source rejection, keepalive, WebSockets and tunnel MTU checked 2026-10-27; Immich not running"
- kind: repository
repository: jupiter-stacks
revision: working-tree
base_revision: 631e4fdfa4c7122e5eff38e11b070c17ff7a4a2a
paths:
- stacks/downloads/docker-compose.yml
- stacks/traefik/docker-compose.yml
- kind: owner-report
reference: "Network topology, Mercury nginx/iptables/fail2ban output, LAN DNS, 2026-10-06"
- kind: owner-report
reference: "Cloudflare ingress YAML and client SSH configurations, 2026-10-06"
- kind: owner-report
reference: "CGNAT rationale, Docker network and container inspection, 2026-10-06"
- kind: owner-report
reference: "Ultra.cc hostname, qBittorrent UI, SFTP rclone remote/systemd unit and Radarr/Sonarr bind mounts, 2026-10-06"
- kind: owner-report
reference: "Authentik fully deployed with all listed app integrations and Authelia retired, 2026-10-16"
- kind: repository
repository: jupiter-stacks
revision: 610e325ef6a98850511ce7a089b0b9a77bfecf15
paths:
- stacks/traefik/docker-compose.yml
- stacks/cloudflared/docker-compose.yml
- stacks/gitea/docker-compose.yml
- stacks/jellyfin/docker-compose.yml
- stacks/audiobookshelf/docker-compose.yml
- kind: repository
repository: jupiter-stacks
revision: d09df456ad331b8beb9a40615114ea8aae2dced5
paths:
- stacks/authentik/docker-compose.yml
- stacks/authentik/blueprints/jupiter.yaml
related: [home-server.reference, home-server.host, home-server.mercury, home-server.downloads, home-server.seedbox, home-server.authentication, home-server.operations]
update_triggers: [dns-change, proxy-change, tunnel-change, firewall-change, certificate-change, seedbox-change]
unknowns:
- public-dns-record-types-and-cloudflare-proxy-status
- cloudflare-access-applications-and-login-policies
- wireguard-allowedips-and-keepalive
- jupiter-firewall-policy
- traefik-forwarded-header-trust-configuration
- complete-airvpn-failure-dns-and-ipv6-isolation-results
revision: 12fb1b96e01fc062fdce4c216f11eb5d7ce716ed
paths: [stacks/traefik/docker-compose.yml, stacks/cloudflared/docker-compose.yml, mercury/nginx/conf.d/upstreams.conf, mercury/nginx/umbra.mom, mercury/nginx/john.gatward.dev, mercury/nginx/samstoreymusic.com]
related: [home-server.mercury, home-server.downloads, home-server.authentication, home-server.deployment]
update_triggers: [dns-change, proxy-change, tunnel-change, firewall-change, certificate-change]
unknowns: [public-dns-record-types-and-cloudflare-proxy-status, cloudflare-access-policies, wireguard-allowedips-and-keepalive, jupiter-firewall-policy]
---
# Networking
## AirVPN downloads
[VPN-isolated local downloads](downloads.md) are live by owner report on
2026-10-26. Gluetun connects directly to AirVPN through the home ISP, separate
from the existing Mercury tunnel. Ultra.cc and the rclone/FUSE mount are retired.
This cuts subscription costs and removes the remote-storage/import dependency.
Full failure, DNS and IPv6 isolation checks are still outstanding.
Public monitoring at `qbittorrent.umbra.mom` is working: Mercury's wildcard
HTTPS ingress reaches Traefik, which joins `downloads` and applies Authentik
per-user/MFA protection before proxying to `gluetun:8080`. The outpost callback
hostname is registered separately. This does not route torrent egress through
Mercury; see the [public-monitoring deployment steps](downloads.md#public-monitoring).
## HTTPS ingress / local downloads
## Application ingress
```mermaid
---
config:
themeCSS: |
.node rect { rx: 8px; ry: 8px; stroke-width: 1.5px; }
.cluster rect { fill: var(--infra-zone) !important; stroke: var(--infra-border) !important; stroke-dasharray: 5 5; rx: 12px; ry: 12px; }
.label, .nodeLabel, .cluster-label { color: var(--infra-text) !important; }
.edgeLabel, .edgeLabel p { background-color: var(--infra-zone) !important; color: var(--infra-text) !important; }
.flowchart-link { stroke: var(--infra-line) !important; }
marker path { fill: var(--infra-line) !important; stroke: var(--infra-line) !important; }
.infraClient rect { fill: var(--infra-client-fill) !important; stroke: var(--infra-client-stroke) !important; }
.infraEdge rect { fill: var(--infra-edge-fill) !important; stroke: var(--infra-edge-stroke) !important; }
.infraService rect { fill: var(--infra-service-fill) !important; stroke: var(--infra-service-stroke) !important; }
flowchart:
curve: basis
nodeSpacing: 35
rankSpacing: 55
text,.nodeLabel,.cluster-label { fill:var(--infra-text);color:var(--infra-text); }
.edgeLabel,.edgeLabel p { background:var(--infra-zone);color:var(--infra-text); }
.flowchart-link { stroke:var(--infra-line); }
marker path { fill:var(--infra-line);stroke:var(--infra-line); }
.cluster rect { fill:var(--infra-zone);stroke:var(--infra-border);stroke-dasharray:5 5;rx:12px;ry:12px; }
.cluster span,.cluster text { color:var(--infra-text)!important;fill:var(--infra-text)!important; }
.infraClient rect { fill:var(--infra-client-fill);stroke:var(--infra-client-stroke);rx:8px;ry:8px; }
.infraEdge rect { fill:var(--infra-edge-fill);stroke:var(--infra-edge-stroke);rx:8px;ry:8px; }
.infraService rect { fill:var(--infra-service-fill);stroke:var(--infra-service-stroke);rx:8px;ry:8px; }
---
flowchart TB
Internet["<b>Public clients</b>"]
subgraph VPS["MERCURY / PUBLIC EDGE"]
Nginx["<b>nginx</b><br/>185.230.217.66<br/>WireGuard 10.0.0.1"]
Uptime["<b>Uptime service</b><br/>127.0.0.1:3001"]
Nginx -->|"uptime.umbra.mom / HTTP"| Uptime
accTitle: Public and LAN application ingress
accDescr: Public HTTPS terminates at Mercury; HTTP crosses encrypted WireGuard to Jupiter. LAN HTTPS bypasses Mercury.
Public["Public clients"]:::infraClient
LAN["LAN clients"]:::infraClient
subgraph Mercury["MERCURY / PUBLIC EDGE"]
Edge["nginx<br/>185.230.217.66<br/>WG 10.0.0.1"]:::infraEdge
Kuma["Uptime Kuma<br/>127.0.0.1:3001"]:::infraService
end
subgraph Home["JUPITER / HOME SERVER"]
LAN["<b>LAN clients</b><br/>Jellyfin / Bookshelf local DNS"]
Traefik["<b>Traefik</b><br/>192.168.1.56<br/>WireGuard 10.0.0.2"]
Apps["<b>Application containers</b><br/>Docker network: traefik"]
Arr["<b>Radarr / Sonarr</b><br/>Download control / media imports"]
Qbit["<b>qBittorrent</b><br/>Shared Gluetun namespace"]
Gluetun["<b>Gluetun</b><br/>WireGuard / firewall<br/>downloads Docker network"]
Downloads["<b>Local torrents / libraries</b><br/>/data/jellyfin<br/>Hardlink imports"]
Router["<b>Home router / ISP</b><br/>Encrypted VPN underlay"]
LAN -->|"HTTPS :443"| Traefik
Traefik -->|"Application routing"| Apps
Traefik -->|"Media management UIs"| Arr
Traefik -->|"Authentik + qBittorrent login / :8080"| Gluetun
Gluetun -->|"Web UI / API"| Qbit
Arr -->|"downloads network / gluetun:8080"| Gluetun
Arr -->|"Local hardlink imports"| Downloads
Qbit -->|"Writes downloads"| Downloads
Qbit -->|"VPN-only torrent traffic"| Gluetun
Gluetun -->|"Encrypted AirVPN WireGuard"| Router
subgraph Jupiter["JUPITER / HOME"]
Proxy["Traefik<br/>LAN 192.168.1.56<br/>WG 10.0.0.2"]:::infraEdge
Apps["Application containers"]:::infraService
end
AirVPN["<b>AirVPN</b><br/>Public VPN exit / reserved peer port"]
Peers["<b>Torrent peers</b>"]
Internet -->|"HTTPS :443"| Nginx
Nginx -->|"HTTPS :443 over WireGuard"| Traefik
Router -->|"Home ISP / not Mercury"| AirVPN
AirVPN <-->|"Torrent traffic"| Peers
class Internet,LAN,Peers infraClient
class Nginx,Traefik,Router,AirVPN infraEdge
class Apps,Uptime,Arr,Gluetun,Qbit,Downloads infraService
Public -->|"HTTPS :443"| Edge
Edge -->|"HTTP :8081 / encrypted WireGuard"| Proxy
LAN -->|"HTTPS :443"| Proxy
Proxy -->|"HTTP / Docker networks"| Apps
Edge -->|"uptime.umbra.mom / HTTP"| Kuma
```
## Cloudflare Tunnel
- Cutover verified by owner **2026-10-27**; browser/OIDC URLs remain HTTPS.
- CGNAT requires Mercury's public IPv4; no residential public ingress.
- Admin hostnames are still public: the tunnel listener is **not private admin access**.
- Torrent egress uses a separate path: [AirVPN](downloads.md#current-topology).
### Listener and trust boundaries
| Traefik entrypoint | Behaviour |
| --- | --- |
| `http`, TCP 80 | Redirect to HTTPS; not Mercury's upstream |
| `https`, TCP 443 | TLS + `letsencrypt` defaults; LAN access and transport rollback |
| `wireguard`, `10.0.0.2:8081` | HTTP; `mercury-only@docker` admits TCP source `10.0.0.1/32` only |
- Public routers: `entrypoints=https,wireguard`; **no router-level TLS settings**.
- Local-only routers: HTTPS only, explicit TLS. Separate HTTPS/WireGuard fallbacks redirect unmatched non-apex hosts to `https://umbra.mom` at priority `1`.
- Forwarded headers trusted only from `10.0.0.1/32` on `https` and `wireguard`.
- nginx sets `Host`, `X-Forwarded-Host`, `X-Forwarded-Port`, `X-Forwarded-Proto`, `X-Real-IP` and `X-Forwarded-For`; client IP is `$remote_addr`, not an appended untrusted chain.
- Admission uses connection source, **not** `X-Forwarded-For`; forged headers do not bypass the allowlist.
- Security-header middleware: `sslProxyHeaders.X-Forwarded-Proto=https` preserves HSTS on the HTTP hop.
- Docker provider: opt-in labels, network `traefik`; file provider: `/data/traefik/dynamic`; custom certificates: `/data/traefik/certs`.
- ACME: Cloudflare DNS-01; `/data/traefik/acme/acme.json`; `umbra.mom` + `*.umbra.mom`.
- Authentication is per application/router, not a universal gate: [Authentication](authentication.md).
### nginx upstream
- Shared `jupiter_backend`: `10.0.0.2:8081`; HTTP/1.1; **16 cached idle connections per worker**, not a concurrency limit.
- `Connection` cleared for normal requests and registry `/v2/`; `upgrade` only for WebSockets.
- Config: `/etc/nginx/conf.d/upstreams.conf` + three domain files in `sites-available/`; `nginx.conf` must include `conf.d/*.conf` and `sites-enabled/*` inside `http`.
- Gitea `/v2/`: unlimited body; `900s` proxy timeouts.
- MTU **1420** on both WireGuard interfaces; full-size non-fragmenting probes passed, approximately **7-8 ms** RTT. No sustained throughput/CPU benchmark.
- Rollout and rollback: [Deployment](deployment.md#ingress-transport-deployment-and-rollback).
## SSH / Cloudflare Tunnel
```mermaid
---
config:
themeCSS: |
.node rect { rx: 8px; ry: 8px; stroke-width: 1.5px; }
.cluster rect { fill: var(--infra-zone) !important; stroke: var(--infra-border) !important; stroke-dasharray: 5 5; rx: 12px; ry: 12px; }
.label, .nodeLabel, .cluster-label { color: var(--infra-text) !important; }
.edgeLabel, .edgeLabel p { background-color: var(--infra-zone) !important; color: var(--infra-text) !important; }
.flowchart-link { stroke: var(--infra-line) !important; }
marker path { fill: var(--infra-line) !important; stroke: var(--infra-line) !important; }
.infraClient rect { fill: var(--infra-client-fill) !important; stroke: var(--infra-client-stroke) !important; }
.infraCloud rect { fill: var(--infra-cloud-fill) !important; stroke: var(--infra-cloud-stroke) !important; }
.infraEdge rect { fill: var(--infra-edge-fill) !important; stroke: var(--infra-edge-stroke) !important; }
.infraService rect { fill: var(--infra-service-fill) !important; stroke: var(--infra-service-stroke) !important; }
flowchart:
curve: basis
nodeSpacing: 35
rankSpacing: 45
text,.nodeLabel { fill:var(--infra-text);color:var(--infra-text); }
.edgeLabel,.edgeLabel p { background:var(--infra-zone);color:var(--infra-text); }
.flowchart-link { stroke:var(--infra-line); }
marker path { fill:var(--infra-line);stroke:var(--infra-line); }
.infraClient rect { fill:var(--infra-client-fill);stroke:var(--infra-client-stroke);rx:8px;ry:8px; }
.infraCloud rect { fill:var(--infra-cloud-fill);stroke:var(--infra-cloud-stroke);rx:8px;ry:8px; }
.infraEdge rect { fill:var(--infra-edge-fill);stroke:var(--infra-edge-stroke);rx:8px;ry:8px; }
.infraService rect { fill:var(--infra-service-fill);stroke:var(--infra-service-stroke);rx:8px;ry:8px; }
---
flowchart TB
Shell["<b>SSH client</b><br/>Host jupiter / user jay"]
Git["<b>Git over SSH</b><br/>Host git.umbra.mom / user git"]
Cloud["<b>Cloudflare Tunnel edge</b><br/>Access policy unconfirmed"]
subgraph Home["JUPITER / HOME SERVER"]
Connector["<b>cloudflared</b><br/>Locally configured ingress"]
SSH["<b>Host sshd</b><br/>192.168.1.56:22"]
Gitea["<b>Gitea SSH</b><br/>192.168.1.56:2222"]
Connector -->|"ssh.umbra.mom / SSH :22"| SSH
Connector -->|"git.umbra.mom / SSH :2222"| Gitea
end
Shell -->|"cloudflared access ssh / ssh.umbra.mom"| Cloud
Git -->|"cloudflared access ssh / git.umbra.mom"| Cloud
Cloud <-->|"Tunnel traffic / initiated outbound by Jupiter"| Connector
class Shell,Git infraClient
class Cloud infraCloud
class Connector infraEdge
class SSH,Gitea infraService
flowchart LR
accTitle: SSH through Cloudflare, separate from application ingress
accDescr: The client uses cloudflared; Jupiter's outbound connector reaches host SSH or Gitea SSH without Mercury or Traefik.
Client["SSH client<br/>cloudflared access ssh"]:::infraClient
Edge["Cloudflare edge"]:::infraCloud
Tunnel["Jupiter cloudflared<br/>outbound connector"]:::infraEdge
Host["Host sshd<br/>192.168.1.56:22"]:::infraService
Git["Gitea SSH<br/>192.168.1.56:2222"]:::infraService
Client --> Edge <--> Tunnel
Tunnel -->|"ssh.umbra.mom"| Host
Tunnel -->|"git.umbra.mom"| Git
```
Client `cloudflared` transports SSH through Cloudflare; Jupiter's connector opens
the outbound tunnel. Origins still perform SSH authentication. No Mercury,
WireGuard, or Traefik hop on this path; no home-router port forwarding.
| Ingress hostname | Origin | Client SSH user |
| --- | --- | --- |
| `ssh.umbra.mom` | `ssh://192.168.1.56:22` | `jay` |
| `git.umbra.mom` | `ssh://192.168.1.56:2222` | `git` |
| Unmatched | `http_status:404` | N/A |
Both client configurations use `IdentityFile ~/.ssh/id_ed25519`.
- No Mercury, WireGuard, Traefik or home-router port forwarding on this path.
- Connector config: `/data/cloudflared/config.yaml` -> `/etc/cloudflared/config.yaml`; unmatched ingress: `http_status:404`.
- Cloudflare Access login/allow policies remain unconfirmed; SSH origins still authenticate users.
```sshconfig
Host jupiter
@@ -197,106 +133,26 @@ Host git.umbra.mom
ProxyCommand cloudflared access ssh --hostname %h
```
Local tunnel configuration is separate from Cloudflare Access applications.
Access login/provider/allow rules remain unconfirmed; `ProxyCommand` alone
does not establish enforcement.
## DNS
Ingress source: host `/data/cloudflared/config.yaml`, mounted at
`/etc/cloudflared/config.yaml`.
## DNS / routes
| DNS / route | Destination |
| Names | Destination |
| --- | --- |
| Public service ingress | Mercury `185.230.217.66` |
| `jellyfin.local.umbra.mom` | Jupiter `192.168.1.56` |
| `bookshelf.local.umbra.mom` | Jupiter `192.168.1.56` |
| Public application names | Mercury `185.230.217.66` |
| `jellyfin.local.umbra.mom`, `bookshelf.local.umbra.mom` | Jupiter LAN `192.168.1.56`; bypass Mercury |
| `ssh.umbra.mom`, `git.umbra.mom` | Cloudflare Tunnel |
| `uptime.umbra.mom` | Mercury nginx -> `http://127.0.0.1:3001` |
| `qbittorrent.umbra.mom` | Mercury -> Traefik -> Authentik -> Gluetun/qBittorrent `8080`; native login retained |
| `gluetun:8080` | Internal authenticated qBittorrent API on the `downloads` Docker network |
## Download connectivity
## Docker networks
qBittorrent now runs on Jupiter in Gluetun's network namespace. The enabled
Gluetun firewall and qBittorrent's `wg0` binding are configured for VPN-only
torrent egress. AirVPN's reserved peer port is allowed through the VPN firewall,
not published on Jupiter or forwarded on the home router. Working downloads
alone do not demonstrate inbound peer reachability or failure isolation.
2026-10-06 supplied snapshot; use service names, not container IPs.
Sonarr/Radarr and optional Prowlarr direct downloads use `gluetun:8080` with
native qBittorrent credentials, not the public Authentik-protected URL.
Their ordinary internet routes and Prowlarr/FlareSolverr indexer traffic remain
outside AirVPN. The Web UI also has a loopback-only `127.0.0.1:8080` publication
for private SSH forwarding.
Imports now use local storage and hardlinks, not SFTP/FUSE. The former
[Seedbox](seedbox.md) page is historical; its service and mount are removed.
See [Storage](storage.md) for current placement and
[Local downloads](downloads.md) for the runbook and remaining checks.
## Mercury
[Mercury VPS](mercury.md): native nginx; WireGuard server `10.0.0.1/24`;
Jupiter client `10.0.0.2/24`. Hosting, TLS, firewall and fail2ban details are
maintained on that page.
Home ISP uses CGNAT. Mercury supplies a static public IPv4 and forwards over
WireGuard; owner preference is to avoid public ingress via the residential IP.
## Observed Docker networks
Snapshot from supplied inspection, not fixed-address configuration. Container
IPs may change; use Docker service names.
| Network | IPv4 subnet | Internal | Observed role |
| Network | Subnet | Internal | Role |
| --- | --- | --- | --- |
| `traefik` | `172.18.0.0/16` | No | Reverse proxy and application ingress |
| `gitea_network` | `172.23.0.0/16` | No | Gitea, runner, notes builder, Havox sync |
| `traefik` | `172.18.0.0/16` | No | Application ingress |
| `gitea_network` | `172.23.0.0/16` | No | Gitea, runner, builder, Havox |
| `cloudflared_cloudflare` | `172.26.0.0/16` | No | Tunnel connector |
| `gitea_data` | `192.168.96.0/20` | Yes | Gitea and PostgreSQL |
| `paperless_data` | `192.168.112.0/20` | Yes | Paperless, PostgreSQL, Redis |
| `dozzle_dozzle` | `172.27.0.0/16` | Yes | Dozzle and socket proxy |
| `gitea_data` | `192.168.96.0/20` | Yes | Gitea/PostgreSQL |
| `paperless_data` | `192.168.112.0/20` | Yes | Paperless/PostgreSQL/Redis |
| `dozzle_dozzle` | `172.27.0.0/16` | Yes | Dozzle/socket proxy |
Cloudflared's observed `172.26.0.2` matches the supplied proxied SSH login source.
An internal network alone does not isolate a multi-network container from
egress through its other networks. Empty network entries do not establish
whether they are obsolete or safe to remove.
The newer `downloads` bridge is defined with IPv6 disabled and is not internal.
Gluetun, Traefik, Sonarr, Radarr and Prowlarr attach to it; qBittorrent shares
Gluetun's namespace rather than joining independently. Its subnet was not
recorded. This supplements, rather than re-inspects, the older table.
## Jupiter: repository configuration
- Traefik publishes TCP `80,443`; HTTP redirects to HTTPS.
- Docker provider: opt-in `traefik.enable=true`; external network `traefik`.
- File provider: `/data/traefik/dynamic`.
- ACME: Let's Encrypt; Cloudflare DNS-01; `/data/traefik/acme/acme.json`;
apex + `*.umbra.mom` certificate requested.
- HTTPS fallback: priority `1`; unmatched non-apex hosts redirect to `https://umbra.mom`.
- `cloudflared`: `/data/cloudflared` -> `/etc/cloudflared`;
stack-local bridge `cloudflare`; no published ports.
## Authentication routing
[Authentik](authentication.md) serves `auth.umbra.mom` through Traefik to
`authentik-server:9000`; the server joins the external `traefik` network.
The ingress diagrams show transport, not a universal authentication gate.
Native OIDC integrations are listed on the authentication page; Backrest, Notes,
Prowlarr, Radarr, Sonarr and qBittorrent
use Traefik middleware `authentik@docker` and the embedded outpost.
Dozzle (`logs.umbra.mom`) now uses OIDC, not forward-auth.
On the six proxy-protected hostnames, a priority-1000
`/outpost.goauthentik.io/` route goes straight to Authentik without forward-auth
to handle callbacks. Other application requests use their normal routers.
The middleware trusts forwarded headers; this is not evidence that the wider
Mercury/Traefik forwarded-header trust chain is configured.
Authentik also publishes `127.0.0.1:9000` for private host access.
Repository-defined networks (not part of the older observed-network snapshot):
`authentik-data` is internal and connects PostgreSQL, server and worker;
the worker also joins `authentik-egress`. The server's additional `traefik`
attachment means the internal data network is not complete egress isolation.
- Newer networks: [Downloads](downloads.md), [Notes build](notes.md#networks-and-webhook-allowlist), [Authentik](authentication.md#state-secrets); not part of this snapshot.
- `internal` does not block egress through a container's other networks; unused-looking networks are not automatically safe to delete.
+55 -84
View File
@@ -2,112 +2,83 @@
schema_version: 1
id: home-server.notes
type: reference
scope: [jupiter, notes, gitea, webhooks, stack-deployment]
scope: [jupiter, notes, gitea, webhooks]
sensitivity: private-infrastructure
last_reviewed: "2026-10-16"
last_reviewed: "2026-10-27"
sources:
- kind: repository
repository: jupiter-stacks
revision: f39d0f4028f9fe70f838a0ec577d10ab11409f5c
paths:
- stacks/gitea/docker-compose.yml
- stacks/notes/docker-compose.yml
- stacks/notes/listener.py
- stacks/notes/nginx.conf
paths: [stacks/gitea/docker-compose.yml, stacks/notes/docker-compose.yml, stacks/notes/listener.py, stacks/notes/nginx.conf]
- kind: owner-report
reference: "Gitea private-address egress rejection, Docker network overlap failure, network inventory and host routes, 2026-10-16"
related: [home-server.reference, home-server.deployment, home-server.networking, home-server.storage, home-server.authentication]
update_triggers: [notes-stack-change, builder-change, webhook-change, network-change, deployment-workflow-change]
unknowns: [successful-deployment-of-notes-build-network, successful-webhook-redelivery]
reference: "Private-address webhook rejection, subnet overlap and host/network inventory, 2026-10-16"
related: [home-server.deployment, home-server.networking, home-server.storage, home-server.authentication]
update_triggers: [builder-change, webhook-change, network-change]
unknowns: [successful-notes-build-network-deployment, successful-webhook-redelivery]
---
# Notes
# Notes publication
The repository configuration separates building the documentation from serving
it. The current network configuration is intended state, not confirmation that
the failed deployment has been recovered.
Repository intent; recovery from the previously failed deployment is unconfirmed.
## Services and build flow
```mermaid
---
config:
themeCSS: |
text,.nodeLabel { fill:var(--infra-text);color:var(--infra-text); }
.edgeLabel,.edgeLabel p { background:var(--infra-zone);color:var(--infra-text); }
.flowchart-link { stroke:var(--infra-line); }
marker path { fill:var(--infra-line);stroke:var(--infra-line); }
.infraClient rect { fill:var(--infra-client-fill);stroke:var(--infra-client-stroke);rx:8px;ry:8px; }
.infraEdge rect { fill:var(--infra-edge-fill);stroke:var(--infra-edge-stroke);rx:8px;ry:8px; }
.infraStorage rect,.infraStorage path { fill:var(--infra-storage-fill);stroke:var(--infra-storage-stroke); }
.infraService rect { fill:var(--infra-service-fill);stroke:var(--infra-service-stroke);rx:8px;ry:8px; }
---
flowchart LR
accTitle: Signed webhook to built documentation
accDescr: Gitea queues a signed push; a single builder writes the shared static volume; nginx serves it read-only.
Git["Gitea push<br/>main"]:::infraClient
Builder["notes_builder :9000<br/>HMAC check / queue / build"]:::infraEdge
Static[("notes_static<br/>generated site")]:::infraStorage
Web["notes_web :8080<br/>read-only nginx"]:::infraService
Git -->|"Signed POST"| Builder -->|"/site-output"| Static -->|"/usr/share/nginx/html"| Web
```
| Component | Role |
| --- | --- |
| `notes_builder` | Zensical container running `/builder/listener.py`; receives signed Gitea webhooks on port `9000` and builds the `main` branch |
| `notes_web` | Unprivileged nginx serving the generated site on port `8080`; Traefik publishes `https://notes.umbra.mom` behind Authentik forward-auth |
| `notes_static` | Shared named volume: builder writes `/site-output`; nginx mounts `/usr/share/nginx/html` read-only |
Gitea sends a push webhook to `http://notes_builder:9000/`. The listener checks
the `X-Gitea-Signature` HMAC against `WEBHOOK_SECRET`, queues accepted deliveries
and returns HTTP `202`. A single worker clones
`http://gitea:3000/jay/notes.git`, then runs:
- Web: `https://notes.umbra.mom`, Traefik + [Authentik forward-auth](authentication.md#forward-auth).
- Clone source: `http://gitea:3000/jay/notes.git`; requires `zensical.toml`, `pyproject.toml`, `uv.lock`.
- Listener `/builder/listener.py` checks `X-Gitea-Signature` with `WEBHOOK_SECRET`.
- **202 = queued**, not built. Single worker logs `Documentation build succeeded` after generated `index.html`.
- Publication replaces volume contents non-atomically; brief missing files possible. Starting an empty builder does not trigger a build.
```sh
uv run --locked zensical build --clean --config-file zensical.toml
```
The repository must contain `zensical.toml`, `pyproject.toml` and `uv.lock`.
After a successful build with a generated `index.html`, the worker replaces the
contents of `notes_static`. Publication is not atomic: nginx can briefly see
missing files during replacement. HTTP `202` means queued, not build completed;
look for `Documentation build succeeded` in the builder logs.
## Networks and webhook allowlist
| Setting | Repository configuration |
| Setting | Value |
| --- | --- |
| Build network | `notes_build`, owned by the Gitea stack and external to the Notes stack |
| Subnet | `10.90.0.0/29` |
| Builder address | `10.90.0.3` |
| Gitea allowlist | `GITEA__security__ALLOWED_HOST_LIST: "10.90.0.3:9000"` |
| Web network | `notes_web` joins `traefik`; the builder does not |
| Build network | `notes_build`, Gitea-owned / external to Notes; `10.90.0.0/29` |
| Builder IP / webhook | `10.90.0.3`; `http://notes_builder:9000/` |
| Gitea setting | `GITEA__security__ALLOWED_HOST_LIST="10.90.0.3:9000"` |
| Web network | `notes_web` on `traefik`; builder not attached |
Gitea 28 requires an explicit IP allow entry for a private destination.
`notes_builder:9000` alone did not permit the resolved Docker IP. Keep the
network subnet, builder static IP and allowlist consistent; do not broadly allow
all private destinations.
The first proposed subnet, `172.30.90.0/29`, overlapped an Actions network using
`172.30.0.0/16`. The replacement does not overlap the owner-supplied network
inventory or routes. Recheck for collisions before rebuilding on another host.
The build network is not internal: the builder needs outbound access to install
Git, uv and locked Python dependencies.
- Gitea 28 requires resolved private IP allowlisting; hostname alone failed. Do not broadly allow private networks.
- Recheck subnet collisions on recovery; old `172.30.90.0/29` overlapped Actions' `172.30.0.0/16`.
- Build network is non-internal: dependency installation requires outbound access.
## Deployment
Stack definitions follow the [Portainer Git deployment workflow](deployment.md).
Changing this documentation does not deploy either stack.
1. Push stack changes; preserve Portainer values. Copy listener/nginx config into `/data/notes/{listener.py,nginx.conf}`; Git pull does not update binds.
2. Set matching `WEBHOOK_SECRET` in Notes and Gitea's push webhook; retain the internal webhook URL.
3. Redeploy **Gitea first**; confirm `notes_build` has the expected subnet/allowlist.
4. Redeploy Notes outside an active build; preserve `notes_static`, never remove volumes.
5. Wait for healthy Gitea and builder `Listening on 9000`; redeliver the webhook.
6. Confirm 202, build-success log and authenticated updated page.
1. Push the updated Compose definitions to `jupiter-stacks/main`. Preserve the
existing production environment settings in Portainer.
2. On Jupiter, ensure `/data/notes/listener.py` and `/data/notes/nginx.conf` contain
the corresponding repository files. These are read-only bind mounts; pulling
the stack repository does not populate or update them.
3. Set `WEBHOOK_SECRET` in the Notes stack environment and the same secret in the
Gitea Notes repository's push webhook. Keep the URL
`http://notes_builder:9000/`; do not put the secret in Git or these notes.
4. Update/redeploy the **Gitea stack first**. It creates `notes_build` and applies
the IP allow entry. Expect a brief Gitea interruption.
5. Confirm `docker network inspect notes_build` reports `10.90.0.0/29`, then
update/redeploy the **Notes stack**. Avoid interrupting an active build.
Preserve `notes_static`; do not remove volumes.
6. Wait for Gitea to be healthy and for `docker logs notes_builder` to show
`Listening on 9000`. Redeliver the failed push webhook in Gitea.
7. Confirm HTTP `202`, a successful build in the builder logs, and the updated
page through `https://notes.umbra.mom` after authentication.
## Health / troubleshooting
Starting the builder alone does not trigger an initial build. If the static
volume is empty, nginx cannot serve the site until an accepted webhook completes.
## Health and troubleshooting
`notes_web` uses `/` on port `8080` for its container health check. For Homarr on
the shared `traefik` network, use `http://notes_web:8080/`.
The builder has no GET health endpoint or Compose health check; its listener
accepts signed POST requests only. Do not configure an ordinary HTTP GET ping
against `http://notes_builder:9000/`, or attach it to `traefik` just for a ping.
If Notes reports `network notes_build declared as external, but could not be
found`, fix the Gitea network creation failure first. A subnet-overlap failure
requires an unused range, not deletion of unrelated Docker networks. If delivery
is accepted but the site does not update, inspect the builder logs for clone,
dependency installation, build or publication errors.
- Web probe: `http://notes_web:8080/` on `traefik`.
- Builder accepts signed POST only; no GET health endpoint/Compose healthcheck. Do not expose it on Traefik for probing.
- Missing external network: fix Gitea's network creation first. Subnet collision: choose a free range, **not deletion of unrelated networks**.
- Accepted delivery without update: inspect clone/dependency/build/publication errors in builder logs.
+52 -188
View File
@@ -2,201 +2,65 @@
schema_version: 1
id: home-server.operations
type: reference
scope: [jupiter, mercury, design-decisions, monitoring, recovery]
scope: [jupiter, mercury, monitoring, troubleshooting]
sensitivity: private-infrastructure
last_reviewed: "2026-10-26"
last_reviewed: "2026-10-27"
sources:
- kind: owner-report
reference: "Sonarr, Radarr and Prowlarr now have alerting enabled; destination and enabled events not supplied, 2026-10-26"
- kind: owner-report
reference: "AirVPN local downloads live, Ultra.cc/rclone retired; cost saving and simpler architecture; GBP 56.06 three-year plan ends 2029-10-25; imports/hardlinks confirmed, full isolation checks and backups outstanding, 2026-10-26"
- kind: owner-report
reference: "RAID/media tradeoff, CGNAT/public IPv4 rationale, no alerting and running monitoring/log containers, 2026-10-06"
- kind: owner-report
reference: "Outages over one day catastrophic; no independent credentials except restic key, recovery evidence, host-config archive or operational rules, 2026-10-06"
- kind: owner-report
reference: "OIDC and three-group motivation; Authentik fully deployed and Authelia retired; updated six-directory backup plan; independent recovery custody unchanged, 2026-10-16"
- kind: owner-report
reference: "Notifications received in Discord server Jupiter: Beszel in #beszel-alerts, Backrest in #backrest-alerts, and Mercury Uptime Kuma in #uptimekuma-alerts, 2026-10-23"
- kind: owner-report
reference: "Mercury Uptime Kuma only checks WireGuard tunnel health and Jupiter online status, notifying Jay on Discord when down; Kuma per-container monitoring not configured, 2026-10-26"
- kind: owner-report
reference: "Beszel sends Discord alerts for any systemd error or any container down for more than two minutes, 2026-10-26"
- kind: owner-report
reference: "Mercury monitored in Beszel; CPU above 90% for 10 minutes, memory above 90% for 10 minutes and any failed systemd services alerts configured, 2026-10-26"
reference: "24h outage tolerance 2026-10-06; Discord receipt 2026-10-23; Beszel/Kuma/Arr alert configuration 2026-10-26"
- kind: live-inspection
reference: "Traefik 3.7.13 startup health/log timing and subsequent route recovery, 2026-10-27"
- kind: documentation
reference: "Beszel v0.21.0 upstream internal/site/src/lib/alerts.ts and internal/alerts/alerts_container.go: Container Health alerts detect Docker unhealthy status, 2026-10-26"
related: [home-server.reference, home-server.host, home-server.mercury, home-server.downloads, home-server.networking, home-server.storage, home-server.authentication, home-server.backups, home-server.deployment]
update_triggers: [data-priority-change, architecture-change, alerting-change, recovery-test]
unknowns:
- acceptable-data-loss-and-recovery-order-by-service
- data-criticality-outside-the-six-backed-up-directories
- tested-full-host-recovery-order
- monitoring-review-frequency
- other-alert-rules-thresholds-and-exact-uptime-probes
- arr-alert-destinations-and-enabled-events
reference: "Traefik v3.7 pkg/provider/docker/config.go health filtering; Beszel v0.21.0 alerts.ts/alerts_container.go health semantics"
related: [home-server.backups, home-server.networking, home-server.authentication, home-server.downloads, home-server.host]
update_triggers: [alerting-change, outage, recovery-test]
unknowns: [exact-uptime-probes-and-timing, other-alert-thresholds, arr-alert-destinations-and-events, monitoring-review-frequency]
---
# Operations / decisions
## Owner decisions
| Decision | Reason / accepted tradeoff |
| --- | --- |
| RAID 0 for HDD data | Capacity prioritized over disk redundancy; downloaded media can be reacquired |
| Back up important state | Six application directories covered, subject to exclusions; media redundancy is not the objective |
| Mercury public edge | Static public IPv4; home ISP uses CGNAT |
| No residential public ingress | Owner preference to avoid exposing the home IP |
| AirVPN instead of Ultra.cc | Live local downloads reduce cost and simplify architecture: no remote seedbox or SFTP/FUSE import dependency; GBP 56.06 prepaid for three years, expires Thursday, 25 October 2029 |
| Authentik instead of Authelia | Native OIDC rather than only forward-auth; retain forward-auth where needed |
| Application access | 2026-10-26: replace the former Admin/Developers/Clients groups with manual per-user service bindings; keep `jay` as administrator with access to every configured application; repository prepared, deployment unconfirmed |
RAID 0 still stores important state alongside media: either HDD failure loses
the array; recovery depends on working backups. Owner intent is not evidence
that every important dataset is covered.
The Authentik migration is complete by owner report; Authelia is retired.
Exact repository group names, app assignments and MFA policy are in
[Authentication](authentication.md). Group access is not the same as application
administrator privileges or Authentik superuser status.
## Download migration
[Local downloads](downloads.md) are live by owner report on 2026-10-26.
Direct home-ISP egress, both Arr hardlink imports, seeding/removal and the public
UI's Authentik plus qBittorrent login are confirmed. Ultra.cc is
cancelled/expired, retained files are handled and the old mount/service removed.
The subscription saving is approximately GBP 35.31/year before incremental
electricity, compared with the former GBP 4.50/month seedbox.
Remaining operational work: complete peer-ingress, DNS/IPv6 and
VPN-failure/startup/reboot isolation checks, and back up qBittorrent state/defaults
and AirVPN secrets. Normal successful downloads do not prove fail-closed behaviour.
The retired seedbox must not be assumed available for rollback.
# Operations
## Observability
| Facility | Current configuration |
Owner-confirmed configuration; Discord server **Jupiter**, receipt confirmed 2026-10-23.
| Source | Coverage / trigger | Destination |
| --- | --- | --- |
| Beszel hub on Jupiter | Jupiter + Mercury; reported systemd errors / container down >2 min | `#beszel-alerts` |
| Mercury Beszel rules | CPU >90% for 10 min; RAM >90% for 10 min; failed systemd units | `#beszel-alerts` |
| Uptime Kuma on Mercury | WireGuard path and Jupiter availability only; alerts Jay | `#uptimekuma-alerts` |
| Backrest | Notifications enabled; exact hook events unknown | `#backrest-alerts` |
| Sonarr / Radarr / Prowlarr | Settings -> Connect enabled; events/destinations unknown | Not recorded |
| Dozzle | Container logs via Docker socket proxy | Web UI |
- Beszel UI/database stores rules. `Failed Services` detects failed units, not every journal error; `Container Health` detects Docker `unhealthy`. Exact stopped/missing-container rules need inspection.
- Beszel cannot alert when Jupiter/hub is down; Kuma is independent of Jupiter but cannot report its own complete outage.
- Kuma has no application/container probes. Arr notifications do not establish empty-search, rejected-release or stalled-torrent coverage.
- Backup notifications do not prove integrity, service restart or detection of a backup that never ran; see [Backups](backups.md).
### Proposed monitoring additions
- Add authenticated-safe application health probes, not proxy/login-page checks; start at 60s / two retries.
- Test outage **and recovery** delivery during maintenance; account for scheduled backup stops.
- Add external Mercury monitoring and missed-backup detection.
- Remote Docker checks require a secured connection; Mercury's socket cannot inspect Jupiter. Never expose an unauthenticated Docker API just for monitoring.
## Traefik startup troubleshooting
- Symptom: `middleware "security-headers@docker" does not exist` immediately after restart.
- Cause: shared middleware lives on Traefik's Docker labels; the provider filters its container while health is `starting`/unhealthy.
- Observed 2026-10-27: errors **11:41:07 UTC**, first healthy check **11:41:11**; Homarr/Authentik routes recovered.
- Action: inspect health + latest logs, then request affected routes. Persistent errors after healthy status require provider/label investigation.
- Do **not** disable health checks or admit unhealthy services globally.
- Possible fix: move shared middleware to the file provider and update references together; **not implemented**.
- `aliasHeadersStrategy` and encoded-path warnings are separate; review backend compatibility before selecting policies. No warning-related settings changed.
## Outstanding work
| Priority | Work / canonical detail |
| --- | --- |
| Host/container monitoring | Beszel monitors Jupiter and Mercury; existing systemd/container alerts plus Mercury CPU, memory and failed systemd service alerts (owner-confirmed) |
| Container logs | Dozzle with Docker socket proxy present |
| Uptime monitoring | Uptime Kuma on Mercury at `uptime.umbra.mom`; WireGuard tunnel health and Jupiter online status only |
| Automatic alerting | Discord delivery configured for Beszel, Backrest and Uptime Kuma; receipt confirmed by owner. Sonarr, Radarr and Prowlarr alerting enabled by owner report; destinations/events not recorded |
### Arr application alerting
**Sonarr, Radarr and Prowlarr all have alerting enabled**, confirmed by the
owner on 2026-10-26. Notification destinations, enabled events and delivery
confirmation have not been recorded. These settings are application-managed
state, not declared in the Compose files.
Do not infer that this covers searches returning no acceptable releases,
missing media or stalled downloads: those conditions are not equivalent to
application health issues or explicit download failures. Alerting for them
requires separate confirmation of the configured events or additional monitoring.
See [Local downloads / AirVPN](downloads.md#alerting).
### Discord notifications
The owner confirms receiving notifications in the Discord server **Jupiter**
as of 2026-10-23:
| Source | Channel |
| --- | --- |
| Beszel | `#beszel-alerts` |
| Backrest | `#backrest-alerts` |
| Uptime Kuma on Mercury | `#uptimekuma-alerts` |
Mercury's Uptime Kuma alerts **Jay** in `#uptimekuma-alerts` when the WireGuard
tunnel or Jupiter availability checks fail; this scope was confirmed on
2026-10-26. Individual containers and applications are not currently monitored
by Kuma.
Beszel sends alerts to **`#beszel-alerts`** for **any systemd error** or **any
container down for more than two minutes**, as confirmed by the owner on
2026-10-26. These are configured alerts, not proposed additions.
Mercury is also monitored in Beszel, with the following alerts configured by
owner report on 2026-10-26:
| Mercury alert | Trigger |
| --- | --- |
| CPU usage | Above 90% for 10 minutes |
| Memory usage | Above 90% for 10 minutes |
| Failed Services | Any failed systemd service |
These settings are stored in the Beszel hub UI/database, not the agent Compose
file. The failed-services rule covers failed systemd units, not all journal
errors.
Exact Kuma probes and timing, other Beszel rules/thresholds and Backrest hook
conditions have not been recorded. Monitoring of Mercury independent of Jupiter and
missed-backup alerts remain unconfirmed. No webhook URLs or tokens are stored
in these notes.
### Alert coverage and optional additions
Beszel's systemd/container outage alerting above is owner-reported deployed
state. The upstream Beszel `0.21.0` **Container Health** rule specifically checks
Docker `unhealthy` status, and **Failed Services** checks failed systemd units,
not every error-level journal entry. The exact live rules and stopped/missing
container detection have not been independently inspected.
The following are optional additions, not configured Kuma monitors:
1. In Mercury's Uptime Kuma, add an **HTTP(s)** monitor for each important
application's health endpoint, or a **TCP Port** monitor where HTTP is not
available. Enable the existing Discord notification on each monitor.
Start with a 60-second interval and two retries to tolerate brief restarts.
Check the application itself, not just a proxy or authentication login page.
Use a stable health endpoint with expected status/body; do not disable
authentication or expose private services solely for monitoring.
2. For additional container-running-state checks, Kuma's **Docker Container**
monitor needs a Docker host connection to the relevant machine. Mercury's
Docker socket only describes Mercury, not Jupiter, and a Portainer environment
connection does not automatically provide Kuma with Docker access. Remote
Docker monitoring needs a separately secured connection; do not expose an
unauthenticated Docker API or mount Jupiter's Docker socket into Kuma merely
to add application availability checks.
3. During a planned maintenance window, stop a non-critical monitored service,
confirm Jay receives the outage alert, then restart it and confirm recovery.
Account for scheduled backup stops before enabling noisy alerts.
Application probes detect service failures even when Docker still reports a
container as running, but are not an inventory of every container's state.
Beszel on Jupiter cannot deliver alerts while Jupiter or the Beszel hub is down;
keep Mercury's independent tunnel/host checks.
The backup plan stops Authentik, interrupting new OIDC logins and forward-auth
access. Authentik/Wardrowbe restart hooks ignore errors, so backup completion
alone does not establish that authentication or applications recovered.
## Recovery status
| Item | Current status |
| --- | --- |
| Jupiter outage tolerance | At most 24 hours; longer is catastrophic to the owner |
| Recovery time | Not measured; 24-hour objective is not demonstrated |
| Acceptable data loss / RPO | Not defined |
| Independent recovery secrets | Restic encryption key only |
| Backup/integrity evidence | None supplied |
| Restore exercise | Never performed |
| Independent host-config archive | Not maintained |
| Operational rules / rollback policy | Not established |
Backup scope is documented in [Backups](backups.md). The six-directory plan now
includes Authentik identity/MFA state and Wardrowbe data, but does not establish
OS, broader host-configuration or named-volume recovery. There is still no
independent Authentik admin/MFA or stack-secret recovery arrangement.
## Proposed recovery storage
Not implemented:
- Private Git repository: sanitized host configuration and bootstrap instructions.
- Encrypted off-host/offline archive: required credentials and secret-bearing configuration.
- Independent access instructions: account recovery, archive location and decryption.
Access/decryption must work without Jupiter or Portainer. A private Git
repository is not a substitute for protecting secret values. Restic key custody
alone does not provide Google Drive access.
| Recovery | Outage tolerance **24h**, not demonstrated. Define per-service RPO/order and exercise [restores](backups.md#recovery-limits). |
| Startup | Ensure WireGuard owns `10.0.0.2` before Traefik binds; persistent ordering/reboot exercise unverified. |
| VPN isolation | Complete [AirVPN failure/startup/DNS/IPv6 exercises](downloads.md#remaining-validation). |
| Exposure | Private admin ingress is not implemented; public hostnames and Portainer's direct port still exist. |
| Docker logs | Confirm [daemon default activation and container recreation](host.md#docker-log-rotation). |
| Identity | Confirm [per-user bindings and Homarr linking](authentication.md); transport verification did not establish them. |
+30 -62
View File
@@ -2,81 +2,49 @@
schema_version: 1
id: home-server.portainer
type: reference
scope: [jupiter, portainer, deployment, configuration-state]
scope: [jupiter, portainer, configuration-state]
sensitivity: private-infrastructure
last_reviewed: "2026-10-16"
last_reviewed: "2026-10-27"
sources:
- kind: owner-report
reference: "Deployed container, ~/portainer_start.sh and updated Backrest plan, 2026-10-06"
- kind: owner-report
reference: "GitHub main/PAT access, Portainer environment values, manual redeployment and container inspection, 2026-10-06"
- kind: owner-report
reference: "Authentik fully deployed with all listed app integrations, including Portainer, 2026-10-16"
- kind: live-inspection
reference: "Image 2.45.3, mounts, publication, restart policy and routers inspected 2026-10-27"
- kind: repository
repository: jupiter-stacks
revision: d09df456ad331b8beb9a40615114ea8aae2dced5
paths:
- stacks/portainer/portainer_start.sh
- stacks/authentik/blueprints/jupiter.yaml
related: [home-server.reference, home-server.storage, home-server.authentication, home-server.backups, home-server.deployment]
update_triggers: [script-change, portainer-upgrade, state-migration, backup-plan-change]
unknowns: [local-hostname-dns-resolution, successful-state-and-secret-restore]
revision: working-tree
base_revision: 12fb1b96e01fc062fdce4c216f11eb5d7ce716ed
reference: "Image correction 2.45.1 -> 2.45.3 uncommitted at review"
paths: [stacks/portainer/portainer_start.sh]
related: [home-server.authentication, home-server.deployment, home-server.backups]
update_triggers: [script-change, portainer-upgrade, state-migration]
unknowns: [local-hostname-dns-resolution, host-script-and-boot-invocation, successful-state-restore]
---
# Portainer
## Deployment
Managed by handwritten `~/portainer_start.sh`, **not a Compose stack**.
Reference script: [portainer_start.sh](https://github.com/jayo60013/jupiter-stacks).
- Standalone shell script, **not Compose**: `stacks/portainer/portainer_start.sh`; supplied host copy `~/portainer_start.sh`.
- Working tree pins **2.45.3**, matching live; committed `12fb1b9` still pins **2.45.1**. Correct before using a fresh clone.
- Script stops/removes then recreates the container. Host copy/boot invocation not revalidated.
| Setting | Deployed value |
| --- | --- |
| Image | `portainer/portainer-ce:2.45.1` |
| Container | `portainer` |
| Restart policy | `always` |
| Network | External Docker network `traefik` |
| State | Host `/data/portainer` -> container `/data` |
| Docker API | `/var/run/docker.sock` bind mount |
| Host publication | TCP `9123` -> container `9000` |
| HTTPS routers | `portainer.umbra.mom`, `portainer.local.umbra.mom` |
| Traefik backend | HTTP; container port `9000` |
| TLS | `https` entrypoint; `letsencrypt` resolver |
| Image / container | `portainer/portainer-ce:2.45.3` / `portainer` |
| Restart / network | `always` / external `traefik` |
| State | `/data/portainer:/data`; not old `portainer_data` volume |
| Docker API | `/var/run/docker.sock` bind |
| Direct port | `9123:9000`; bypasses Traefik middleware |
| Public router | `portainer.umbra.mom`; `https,wireguard`; no router TLS labels |
| Local router | `portainer.local.umbra.mom`; HTTPS-only, explicit TLS/`letsencrypt` |
| Backend | HTTP `9000`; no forward-auth middleware |
Script stops/removes the existing container, then recreates it; bind-mounted
state persists. Requires Docker, the `traefik` network, and mounted `/data`.
Supplied routers attach no forward-auth middleware. Local router does not
establish local DNS resolution.
- Replace from independent SSH, not Portainer's UI; require mounted `/data`, Docker and `traefik`.
- Back up state first; preserve image/mounts/network/restart/port. Never run two instances against the same state; a stopped container is not a DB backup.
- Shared routing convention: [Networking](networking.md#listener-and-trust-boundaries).
## Authentication
## Persisted state
Native OIDC through [Authentik](authentication.md); client ID `portainer`;
issuer `https://auth.umbra.mom/application/o/portainer/`.
Both public and local HTTPS hostnames are registered as strict callbacks.
The updated Authentik blueprint grants `jay` direct access, with other users
assigned manually per application instead of through the retired `development`
group. Deployment of this access-policy migration is unconfirmed. Admission
does not itself assign Portainer administrator permissions.
OAuth configuration is database/UI state, not configured by the container
startup script. The owner reports the integration complete; local-login and
account-linking settings were not supplied. Preserve `/data/portainer` for
both stack management and authentication configuration.
## Stack management
Portainer fetches <https://github.com/jayo60013/jupiter-stacks>, branch `main`,
using a GitHub PAT. Stack environment values/secrets are managed in Portainer.
Repository changes are pulled/redeployed manually; committing alone does not
update containers. See [deployment](deployment.md).
## Backups
Daily 03:00 UTC; Google Drive; last four snapshots. Backrest reads
`/userdata/portainer` from host `/data/portainer`.
- Before snapshot: stop `portainer`; timeout `120s`; errors fatal.
- Snapshot end/error: start `portainer`; restart errors ignored.
- Restore remains untested; script itself is outside the listed backup paths.
The former `portainer_data` named volume is not used by the deployed container.
- `/data/portainer` holds stack management, environment secrets, Git credentials and database/UI OAuth settings.
- Native OIDC/callbacks and access policy: [Authentication](authentication.md#native-oidc).
- Git pull/redeploy workflow: [Deployment](deployment.md).
- Backup schedule, hooks and restore limits: [Backups](backups.md).
+17 -124
View File
@@ -2,136 +2,29 @@
schema_version: 1
id: home-server.seedbox
type: reference
scope: [seedbox, ultra-cc, qbittorrent, rclone, sftp, media-import]
scope: [retired-seedbox, ultra-cc]
sensitivity: private-infrastructure
last_reviewed: "2026-10-26"
last_reviewed: "2026-10-27"
sources:
- kind: owner-report
reference: "Ultra.cc subscription active from 2024-05-25 until retirement, always GBP 4.50/month with no price changes, 2026-10-26"
- kind: owner-report
reference: "Ultra.cc cancelled/expired, retained files handled, rclone-seedbox service and mount removed; local AirVPN downloads live, 2026-10-26"
- kind: owner-report
reference: "Ultra.cc plan, remaining capacity, rclone remote/process and enabled systemd unit, 2026-10-06"
- kind: owner-report
reference: "apt-managed rclone, active FUSE mount and Radarr/Sonarr bind-mount inspection, 2026-10-06"
- kind: owner-report
reference: "Import-then-delete intent, Remove Completed enabled, sample ratio about 0.01 and global seeding ratio limit 1, 2026-10-06"
- kind: documentation
reference: "https://wiki.servarr.com/radarr/settings#remove-completed-downloads"
- kind: documentation
reference: "https://wiki.servarr.com/sonarr/settings#remove-completed-downloads"
- kind: repository
repository: jupiter-stacks
revision: 610e325ef6a98850511ce7a089b0b9a77bfecf15
paths:
- stacks/radarr/docker-compose.yml
- stacks/sonarr/docker-compose.yml
related: [home-server.reference, home-server.downloads, home-server.networking, home-server.storage, home-server.backups]
update_triggers: [hosting-renewal, quota-change, remote-change, mount-change, media-import-change]
unknowns:
- total-traffic-allowance-and-reset-period
- renewal-behaviour
- rclone-cache-location-and-size-limit
- import-copy-move-settings
- qbittorrent-seeding-limit-action-and-per-torrent-overrides
- download-client-categories-and-api-removal-outcome
- mount-startup-ordering-relative-to-docker-containers
- seedbox-backup-policy
reference: "Former connection snapshot 2026-10-06; cancellation, retained files handled and mount/service removed 2026-10-26"
related: [home-server.downloads]
update_triggers: [archive-correction]
unknowns: []
---
# Seedbox (retired)
# Seedbox archive
**Retired by owner report on 2026-10-26:** Ultra.cc is cancelled/expired,
retained files have been handled, and Jupiter's `rclone-seedbox.service` and
`/mnt/seedbox` FUSE mount are removed. The replacement is
[local qBittorrent through AirVPN](downloads.md), now live with working
Sonarr/Radarr hardlink imports and seeding/removal.
- **Retired 2026-10-26:** Ultra.cc cancelled/expired; retained files handled; rclone unit/FUSE mount removed.
- Replacement, costs and current operations: [Local downloads / AirVPN](downloads.md).
- Historical reference only; **not a rollback or reconstruction target**.
The change reduces subscription cost and simplifies the architecture by removing
remote storage and the SFTP/FUSE import dependency. AirVPN cost GBP 56.06 for
three years, ending **Thursday, 25 October 2029**, versus Ultra.cc's former
GBP 4.50/month.
The Ultra.cc subscription had been active since **2024-05-25**, always at
**GBP 4.50/month**, with no price changes before retirement.
Everything below is a **historical 2026-10-06 snapshot**, retained for reference,
not current service state or a service to recreate.
## Former service / subscription
| Field | Configuration |
| Former component | Value |
| --- | --- |
| Provider / plan | [Ultra.cc](https://ultra.cc/); `Lancer-v2` |
| Host | `terra.usbx.me` |
| Account / home | `jayo60013` / `/home/jayo60013` |
| Cost | GBP 4.50/month |
| Reported expiration | `2026-10-25`; renewal behaviour unconfirmed |
| Disk | 1001 GB allocation; 984 GB available at review |
| Remaining traffic | 2 TB at review; total allowance/reset period unconfirmed |
| Download speed | Provider reports `Unlimited`; no numeric rate supplied |
| Upload speed | Provider reports `50000 Mbps`; not measured throughput |
| qBittorrent UI | <https://jayo60013.terra.usbx.me/qbittorrent/> |
| Provider / plan / host | Ultra.cc `Lancer-v2`; `terra.usbx.me` |
| Remote | rclone `seedbox`, SFTP; `seedbox:downloads/qbittorrent` |
| Mount / consumers | `/mnt/seedbox` -> Arr `/downloads`; remote-to-local hardlinks impossible |
| Removed unit | `/etc/systemd/system/rclone-seedbox.service`; user/group `jay` |
| Former options | `--allow-other --dir-cache-time 1m --vfs-cache-mode writes`; restart on failure, 10s delay |
## Jupiter connection
Actual connection: **SFTP via rclone/FUSE**, not WebDAV or scheduled sync.
| Component | Path / value |
| --- | --- |
| rclone remote | `seedbox`; type `sftp`; host `terra.usbx.me`; user `jayo60013` |
| SSH key file | `/home/jay/.ssh/id_ed25519`; key contents excluded |
| Remote source | `seedbox:downloads/qbittorrent` |
| Remote absolute path | `/home/jayo60013/downloads/qbittorrent` |
| Jupiter mount | `/mnt/seedbox` |
| Container source | `/downloads` in Radarr and Sonarr |
| Radarr library | Container `/movies` -> host `/data/jellyfin/movies` |
| Sonarr library | Container `/tv` -> host `/data/jellyfin/tv_shows` |
rclone remote: `shell_type=unix`; checksum commands `md5sum`, `sha1sum`.
rclone is apt-managed. Observed FUSE mount: UID/GID `1000:1000`;
`rw,nosuid,nodev,allow_other`.
## Mount lifecycle
`/etc/systemd/system/rclone-seedbox.service`; enabled at boot.
User/group `jay`; `Type=simple`; wants/starts after `network-online.target`.
```sh
/usr/bin/rclone mount seedbox:downloads/qbittorrent /mnt/seedbox \
--allow-other \
--dir-cache-time 1m \
--vfs-cache-mode writes
```
Stop: `/bin/fusermount -u /mnt/seedbox`.
Restart: `on-failure`; delay `10s`; target `multi-user.target`.
`--allow-other` permits other local users, subject to filesystem permissions.
`writes` caches write/read-write opens; read-only opens stream from the remote.
No explicit cache path/size limit is supplied in the unit.
The remote mount is not a complete local replica or backup. Radarr/Sonarr
library destinations are on the HDD array; cross-filesystem hardlinks are
impossible. Mount startup ordering relative to Docker containers is unconfirmed.
Their `/downloads` binds use `rprivate`; host remounts are not automatically
propagated into existing containers.
The FUSE-reported capacity is not evidence of the account quota; use the
provider allocation above.
## Import / removal
Intent: Radarr/Sonarr import required media, then remove completed downloads.
Owner reports `Completed Download Handling / Remove Completed` enabled.
qBittorrent global seeding ratio limit: `1.0`.
Reported undeleted example: share ratio approximately `0.01`; seeded less
than one minute. That ratio does not meet the configured global limit.
`Completed` download status alone does not establish completed seeding.
Upstream requirements: successful import, supported seeding goal reached,
client reports seeding complete, paused/stopped torrent, expected category.
Removal is requested through qBittorrent's API, not by deleting through rclone.
Limit action, torrent overrides and removal after reaching the goal remain
unconfirmed; no configuration change or successful-removal evidence recorded.
- This retired unit is unrelated to Backrest's Google Drive/rclone dependency.
+60 -178
View File
@@ -2,174 +2,73 @@
schema_version: 1
id: home-server.storage
type: reference
scope: [jupiter, downloads, disks, raid, filesystems, docker-volumes, media-import]
scope: [jupiter, disks, raid, filesystems, docker-volumes, media-import]
sensitivity: private-infrastructure
last_reviewed: "2026-10-26"
last_reviewed: "2026-10-27"
sources:
- kind: owner-report
reference: "Local AirVPN downloads live; Sonarr/Radarr imports, hardlinks and seeding/removal confirmed; retained seedbox files handled and rclone service/mount removed; download-state backups not yet added, 2026-10-26"
reference: "lsblk/mdadm/fstab/findmnt/volume snapshots 2026-10-06; local hardlinks/imports and remote-mount removal confirmed 2026-10-26"
- kind: repository
repository: jupiter-stacks
revision: working-tree
base_revision: 631e4fdfa4c7122e5eff38e11b070c17ff7a4a2a
paths:
- stacks/downloads/docker-compose.yml
- stacks/sonarr/docker-compose.yml
- stacks/radarr/docker-compose.yml
- kind: owner-report
reference: "lsblk filesystem output, RAID level, mount placement and Docker data root, 2026-10-06"
- kind: owner-report
reference: "Ultra.cc plan/capacity, seedbox rclone mount/systemd unit and Radarr/Sonarr bind-mount inspection, 2026-10-06"
- kind: owner-report
reference: "Deployed Portainer uses /data/portainer bind mount, not portainer_data, 2026-10-06"
- kind: owner-report
reference: "mdstat, mdadm scans/detail, fstab, findmnt and Docker volume/container inspection, 2026-10-06"
- kind: owner-report
reference: "Authentik deployed, Authelia retired; six-directory backup plan includes Authentik and Wardrowbe, 2026-10-16"
- kind: repository
repository: jupiter-stacks
revision: d09df456ad331b8beb9a40615114ea8aae2dced5
paths:
- stacks/backrest/docker-compose.yml
- stacks/jellyfin/docker-compose.yml
- stacks/authentik/docker-compose.yml
- stacks/wardrowbe/docker-compose.yml
- stacks/traefik/docker-compose.yml
related: [home-server.reference, home-server.host, home-server.downloads, home-server.seedbox, home-server.portainer, home-server.backups, home-server.operations]
update_triggers: [disk-change, raid-change, mount-change, volume-change, backup-coverage-change, seedbox-change]
unknowns: [persistent-mdadm-config-and-initramfs-content, disk-health]
paths: [stacks/downloads/docker-compose.yml, stacks/sonarr/docker-compose.yml, stacks/radarr/docker-compose.yml]
related: [home-server.host, home-server.downloads, home-server.backups]
update_triggers: [disk-change, raid-change, mount-change, volume-change]
unknowns: [persistent-mdadm-config-and-initramfs, disk-health]
---
# Storage
[AirVPN/local downloads](downloads.md) are live by owner report on 2026-10-26.
Downloads and libraries now share local storage, with Sonarr/Radarr hardlink
imports and seeding/removal confirmed. Existing host library paths are unchanged.
The Ultra.cc service and rclone/FUSE mount are retired, simplifying imports.
## Physical placement
```mermaid
---
config:
themeCSS: |
.node rect { rx: 8px; ry: 8px; stroke-width: 1.5px; }
.cluster rect { fill: var(--infra-zone) !important; stroke: var(--infra-border) !important; stroke-dasharray: 5 5; rx: 12px; ry: 12px; }
.label, .nodeLabel, .cluster-label { color: var(--infra-text) !important; }
.edgeLabel, .edgeLabel p { background-color: var(--infra-zone) !important; color: var(--infra-text) !important; }
.flowchart-link { stroke: var(--infra-line) !important; }
marker path { fill: var(--infra-line) !important; stroke: var(--infra-line) !important; }
.infraStorage rect, .infraStorage path { fill: var(--infra-storage-fill) !important; stroke: var(--infra-storage-stroke) !important; }
.infraCloud rect { fill: var(--infra-cloud-fill) !important; stroke: var(--infra-cloud-stroke) !important; }
.infraClient rect { fill: var(--infra-client-fill) !important; stroke: var(--infra-client-stroke) !important; }
.infraService rect { fill: var(--infra-service-fill) !important; stroke: var(--infra-service-stroke) !important; }
flowchart:
curve: basis
nodeSpacing: 30
rankSpacing: 45
text,.nodeLabel,.cluster-label { fill:var(--infra-text);color:var(--infra-text); }
.flowchart-link { stroke:var(--infra-line); }
marker path { fill:var(--infra-line);stroke:var(--infra-line); }
.cluster rect { fill:var(--infra-zone);stroke:var(--infra-border);stroke-dasharray:5 5;rx:12px;ry:12px; }
.cluster span,.cluster text { color:var(--infra-text)!important;fill:var(--infra-text)!important; }
.infraStorage rect,.infraStorage path { fill:var(--infra-storage-fill);stroke:var(--infra-storage-stroke); }
.infraEdge rect { fill:var(--infra-edge-fill);stroke:var(--infra-edge-stroke);rx:8px;ry:8px; }
.infraService rect { fill:var(--infra-service-fill);stroke:var(--infra-service-stroke);rx:8px;ry:8px; }
---
flowchart LR
subgraph Import["JUPITER / LOCAL MEDIA IMPORT"]
QBIT["<b>qBittorrent</b><br/>Gluetun / AirVPN"]
TORRENTS[("<b>/data/jellyfin/torrents</b><br/>incomplete / movies / tv")]
ARR["<b>Radarr / Sonarr</b><br/>/data/jellyfin mounted at /media"]
LIBRARY[("<b>/data/jellyfin</b><br/>movies / tv_shows libraries")]
QBIT -->|"Writes /media/torrents"| TORRENTS
TORRENTS -->|"Local download source"| ARR
ARR -->|"Hardlink imports / same filesystem"| LIBRARY
flowchart TB
accTitle: SSD system storage and non-redundant HDD application storage
accDescr: Docker volumes and caches are on SSD; service bind mounts, downloads and libraries are on the striped HDD array.
subgraph SSD["SSD / SAMSUNG 860 EVO / 1 TB"]
Root["sdc3 / ext4 root"]:::infraStorage
Docker["/var/lib/docker<br/>/var/lib/containerd"]:::infraService
Cache["/home/jay/data<br/>service caches"]:::infraService
Root --> Docker
Root --> Cache
end
subgraph Array["HDD ARRAY / APPLICATION DATA"]
direction TB
A["<b>Seagate BarraCuda</b><br/>sda1 / 2 TB"]
B["<b>Seagate BarraCuda</b><br/>sdb1 / 2 TB"]
RAID["<b>Linux MD RAID 0</b><br/>md0 / no redundancy"]
DATA[("<b>/data</b><br/>ext4")]
A --> RAID
B --> RAID
RAID --> DATA
subgraph HDD["HDD / 2 x SEAGATE BARRACUDA / 2 TB EACH"]
Disks["sdb1 + sda1"]:::infraStorage
RAID["md0 / RAID 0<br/>3.64 TiB / no redundancy"]:::infraEdge
Data[("/data / ext4<br/>bind mounts + media")]:::infraStorage
Disks --> RAID --> Data
end
subgraph System["SSD / SYSTEM"]
direction TB
SSD["<b>Samsung 860 EVO</b><br/>sdc / 1 TB"]
EFI["<b>/boot/efi</b><br/>sdc1 / FAT32"]
SWAP["<b>Swap</b><br/>sdc2"]
ROOT["<b>/</b><br/>sdc3 / ext4"]
DOCKER["<b>/var/lib/docker</b><br/>Named volumes"]
CACHE["<b>/home/jay/data</b><br/>Backrest / Jellyfin caches"]
SSD --> EFI
SSD --> SWAP
SSD --> ROOT
ROOT --> DOCKER
ROOT --> CACHE
end
DATA -->|"Backs torrent storage"| TORRENTS
DATA -->|"Backs organised libraries"| LIBRARY
class A,B,SSD,DATA,ROOT,TORRENTS,LIBRARY infraStorage
class RAID infraCloud
class EFI,SWAP infraClient
class DOCKER,CACHE,QBIT,ARR infraService
```
RAID 0: striping; no redundancy; either HDD failure loses the array.
- Either HDD failure loses the array, including application state; [backup coverage](backups.md#coverage) is explicit, not all of `/data`.
- SSD cache paths: `/home/jay/data/{backrest/cache,jellyfin/cache,wardrowbe/ollama,wardrowbe/rembg}`.
- `/data/<service>` holds service bind mounts; application-specific mappings live on each service page.
- Wardrowbe binds: `/data/wardrowbe/{postgres,uploads,redis}`; SSD models/caches above.
## Local downloads / imports
## RAID / mounts
The current path is entirely on Jupiter's HDD array:
```text
qBittorrent -> /data/jellyfin/torrents/{incomplete,movies,tv}
-> Sonarr/Radarr via /data/jellyfin:/media
-> hardlinks in /data/jellyfin/{movies,tv_shows}
```
Sonarr's library root is `/media/tv_shows`; Radarr's is `/media/movies`.
qBittorrent receives only the torrents subtree at `/media/torrents`, not the
organised libraries. Jellyfin retains `/data/jellyfin:/media`; its library roots
must exclude the torrent folders.
Hardlinks let imported files remain available while torrents seed without a
second full data copy. Removing one name does not remove the data while another
link remains; hardlinks are not independent copies or backups. The old remote
mount could not support cross-filesystem hardlinks to the local libraries.
The six-directory Backrest plan does not cover media, torrent data, qBittorrent
state/defaults or VPN secrets. The owner confirms the latter recovery gap
remains. Historical remote storage details are on [Seedbox (retired)](seedbox.md);
retained files are handled and its mount/service are removed.
## Filesystems
Device names reflect the supplied snapshot; UUIDs identify filesystems.
| Mount | Device | Filesystem | UUID |
| --- | --- | --- | --- |
| `/data` | `/dev/md0` | ext4 | `d60390a1-7db3-44cc-bbde-693c23e42001` |
| `/` | `/dev/sdc3` | ext4 | `7a5e0ee6-4a65-4f68-9128-b3b4f3406447` |
| `/boot/efi` | `/dev/sdc1` | FAT32 | `1DDE-CDF9` |
| swap | `/dev/sdc2` | swap | `4781ea5d-fba5-438d-8d97-6a0801669ca1` |
MD members: `/dev/sda1`, `/dev/sdb1`; metadata `1.2`; array label `jupiter:0`;
MD UUID `c67fccaa-e541-bf2d-2bf6-83221086e0e9`.
## RAID assembly / geometry
Linux MD, inspected with mdadm; ext4 directly on the array, no intervening LVM.
| Setting | Observed value |
| Setting | Snapshot value |
| --- | --- |
| Device / scan name | `/dev/md0` / `/dev/md/0` |
| mdadm UUID | `c67fccaa:e541bf2d:2bf68322:1086e0e9` |
| Capacity | 3.64 TiB / 4.00 TB |
| Layout / chunk | `original` / 512 KiB |
| Member order | Slot 0: `/dev/sdb1`; slot 1: `/dev/sda1` |
| State at review | `clean`; 2 active members; no failed/spare devices |
Scan output identifies the array; persisted mdadm configuration/initramfs were
not supplied. Assembly is not array creation; do not substitute `mdadm --create`.
## Mount configuration
| Array | Linux MD `/dev/md0` (`/dev/md/0` scan name); ext4 directly, no LVM |
| MD UUID / label | `c67fccaa:e541bf2d:2bf68322:1086e0e9` / `jupiter:0` |
| Metadata / layout / chunk | `1.2` / `original` / 512 KiB |
| Member order | Slot 0 `/dev/sdb1`; slot 1 `/dev/sda1` |
| State | Clean, 2 active, no failed/spare members at inspection |
| `/data` options | `rw,relatime,stripe=256`; `/tmp` is tmpfs |
| SSD partitions | `sdc1` EFI; `sdc2` swap; `sdc3` root |
Supplied `/etc/fstab`:
@@ -180,50 +79,33 @@ UUID=d60390a1-7db3-44cc-bbde-693c23e42001 /data ext4 defaults 0 2
UUID=4781ea5d-fba5-438d-8d97-6a0801669ca1 none swap sw 0 0
```
Observed `/data`: `rw,relatime,stripe=256`; `/tmp`: tmpfs.
The former seedbox FUSE mount was service-managed, not listed in fstab; it has
now been removed.
- Device names are snapshots; UUIDs identify filesystems.
- Persisted mdadm/initramfs configuration unverified. **Assembly is not creation: do not use `mdadm --create`.**
## Placement
## Media mounts
| Host path | Backing / purpose |
| --- | --- |
| `/data/<service>` | HDD array; service bind mounts |
| `/data/jellyfin/torrents/{incomplete,movies,tv}` | HDD array; local qBittorrent downloads |
| `/data/jellyfin/{movies,tv_shows}` | HDD array; organised media libraries; hardlink imports |
| `/data/qbittorrent/{config,defaults}` | HDD array; client state and startup defaults/scripts; not backed up |
| `/data/gluetun/secrets` | HDD array; root-restricted AirVPN WireGuard secrets; not backed up |
| `/home/jay/data/backrest/cache` | SSD; Backrest cache |
| `/home/jay/data/jellyfin/cache` | SSD; Jellyfin cache |
| `/var/lib/docker` | SSD; Docker state and default local named volumes |
| `/var/lib/containerd` | SSD; observed containerd overlayfs snapshot storage |
| `/data/portainer` | HDD array; deployed [Portainer](portainer.md) state; backed up |
| `/data/authentik/{postgres,data,blueprints}` | Authentik PostgreSQL, persistent files and custom blueprint; backed up |
| `/data/wardrowbe/{postgres,uploads}` | Wardrowbe database and uploads; backed up |
| `/data/wardrowbe/redis` | Wardrowbe Redis; excluded from backups |
| `/home/jay/data/wardrowbe/{ollama,rembg}` | SSD; Wardrowbe models/caches; outside backup plan |
| `/data/traefik/{acme,dynamic,certs}` | ACME state, dynamic configuration, certificates |
| Consumer | Host -> container | Paths |
| --- | --- | --- |
| qBittorrent | `/data/jellyfin/torrents:/media/torrents` | `incomplete`, `movies`, `tv`; no organised-library mount |
| Sonarr / Radarr | `/data/jellyfin:/media` | Roots `/media/tv_shows`, `/media/movies`; categories `sonarr` -> torrents `tv`, `radarr` -> torrents `movies` |
| Jellyfin | `/data/jellyfin:/media` | Library roots exclude `torrents` |
Cache placement is per mount, not globally SSD-backed. `/data` backups do not
implicitly include SSD-backed Docker volumes.
Authelia is retired; retention/deletion of its former `/data/authelia` directory
is unconfirmed. Authentik uses PostgreSQL, not the former Authelia SQLite store.
- Same-filesystem hardlink imports confirmed; no second full copy while seeding.
- Unlinking one name preserves data while another link exists; hardlinks are **not backups**.
- Download service/configuration: [AirVPN](downloads.md). Former FUSE mount is retired.
## Named-volume snapshot
## Named volumes
Inspected volumes below use driver `local`, no driver options, and
`/var/lib/docker/volumes/<name>/_data` on SSD.
Snapshot: driver `local`, no options; `/var/lib/docker/volumes/<name>/_data` on SSD.
| Volume | Running consumer / container path |
| Volume | Consumer path |
| --- | --- |
| `audiobookshelf_config` | Audiobookshelf `/config` |
| `jellyfin_config` | Jellyfin `/config` |
| `notes_static` | Notes web `/usr/share/nginx/html`; builder `/site-output` |
| `notes_static` | Builder `/site-output`; web `/usr/share/nginx/html` |
| `dozzle_dozzle_data` | Dozzle `/data` |
| `ctq_ctq_postgres_data` | CTQ PostgreSQL `/var/lib/postgresql/data` |
| `havox_src` | Havox web/sync `/git` |
Other volumes are present without a running consumer shown, including
`jellyfin_cache`, `notes_notes_static`, and monitoring volumes. Presence does
not prove use or justify deletion. Named volumes are outside the supplied
six-directory backup plan.
- `jellyfin_cache`, `notes_notes_static` and monitoring volumes had no shown running consumer; **not evidence they are safe to delete**.
- Named-volume recovery is separate from `/data` bind mounts: [Backups](backups.md#recovery-limits).