Webhook authenticated and triggered a successful DeployStack (komodo-core
log 18:39:00). Only remaining step is wiring internal jobs as checks.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Stack deployed to Komodo (id 6a3acf2ca7867a4fbab9bfc1), both containers
healthy, Traefik route + LE cert OK, DNS resolves, superuser created and
auth-verified. Flip status to live in ARCHITECTURE 7.6, SERVICE_CATALOG,
MASTER_TODO and the stack README. Document the new host secret files
(secret_key, superuser_password = login password, webhook_secret) in
SECRETS_MAP. Remaining operator step: the Gitea->Komodo webhook.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Codex drills proved all six alert paths to the phone (send-ntfy, restore wrapper, freshness negative, Alertmanager->bridge->ntfy, docker-critical-watcher smoke, Borg pre-hook failure). Add a Kurzlog entry, note the send-ntfy exec-bit beleg-bug (6870ae5), and the one optional gap (Prometheus->AM leg via temp rule). Trim oldest entry to stay at max 5.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
run-restore-job-with-ntfy.sh execs send-ntfy.sh directly; without the exec bit the failure-alert path errored with Permission denied (found during Codex alert drill 2026-06-23). Set the exec bit in the repo to match the live fix.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
backend_net was recreated with --internal (Codex live): egress from postgresql17 blocked, all 12 members reattached, frontends and DB connections verified. Move the parked #17 item to the MASTER_TODO Kurzlog and confirm the live state in NETWORK_INVENTORY. No dawarich_egress needed (sidekiq makes no external connections).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Capture the audit egress analysis durably so the deferred maintenance window keeps the prep. backend_net -> internal:true is the only remaining P3 item; the single risk is dawarich_sidekiq (the only backend_net-only worker), all DB/cache and dual-homed containers are safe. If sidekiq needs egress, use a dedicated dawarich_egress net (immich_egress precedent).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Codex first live run passed (SUCCESS, 7 snapshots, single-file restore from .zfs/snapshot; report hetzner-snapshot-2026-06-23.md) with no ENV overrides. Set runbook status to active, document the run, and add the monthly cadence (15th, cron 0 6 15 * *) to schedule.md and the restore-tests README. Remaining host step: create the Unraid User Script restore-hetzner-snapshot-monthly.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add an Aktiv row for the remaining Codex/operator follow-ups (#19 snapshot-test validation, #12 default-bridge recreate, #14 drift redeploy, #15 immich path) and a Kurzlog entry summarizing the closed P1/P2 core (Vault /admin + Komodo de-publicized, snapshots proven, auth-matrix). Trim oldest Kurzlog entry to stay at max 5.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Make the off-site snapshot protection a repeatable, monitored proof (DECISIONS 2026-06-11/-23): a read-only restore-test that lists .zfs/snapshot on the Storage Box, checks retention and newest-snapshot age, and SFTP-fetches one small file from the newest snapshot (size + SHA256). Connection is derived from the borg-ui repo URL and runs via docker exec borg-ui; no secret in the script, no write access. Wired into the run-restore-checks.sh dispatcher; runbook documents the pending one-time live validation.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Codex applied the ipallowlist middleware (Tailnet 100.64.0.0/10 + LAN 192.168.178.0/24) to the Komodo router live in the inline-managed self-stack; public now returns 403. Mirror the labels in ops/komodo/docker-compose.yml for parity (not auto-deployed), record the decision in docs/DECISIONS.md, and update docs/AUTH_MATRIX.md.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Consolidate effective access policy per public domain (Authelia bypass/two_factor, native exceptions, Tailscale-only, IP-allowlist) into a single reviewable matrix, surfacing the Authelia bypass list that previously lived only in the live config. Indexed in docs/README.md.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Audit 2026-06-23 (P1): /admin was publicly reachable (200). Add a higher-priority Traefik router scoped to PathPrefix(/admin) with an ipallowlist middleware (Tailnet 100.64.0.0/10 + LAN 192.168.178.0/24); the main router stays native for browser and mobile clients. Documented in docs/DECISIONS.md.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
dawarich_redis was the last redis instance still on 7-alpine; the
closed PR #10 kept it as an "Ignored or Blocked" entry in the Renovate
Dependency Dashboard (issue #6). Bump to the already-running
redis:8.8.0-alpine digest and add apps/dawarich to the renovate redis
8.x allowedVersions pin. Data path /mnt/user/appdata/dawarich/redis
unchanged; redis 8 loads the existing RDB snapshots.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bisher war die H:-Nearline-Kopie nur als Backup-Ziel beschrieben
(CAPACITY_AND_LIFECYCLE), nicht als Restore-Quelle. Im Ernstfall fehlte der
Hinweis, dass auf baerchen eine frische lokale Kopie aller Dumps + Bundles
liegt.
- ops/h-drive-nearline/README.md: neuer Abschnitt "Restore aus H:/ (DR-Fall)"
mit Inhalt, Rueckspiel-Weg (-> RESTORE_MATRIX / SERVICES_RECOVERY) und
Pflicht-Frische-Pruefung (deckt den S4U-Stale-Fall ab).
- docs/DISASTER_RECOVERY.md: baerchen-Abschnitt verweist jetzt auf die
H:-Fallback-Restore-Quelle und das Runbook.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Task "KalliLab H Drive Nearline Pull" am 2026-06-21 von S4U auf LogonType
Interactive ("Nur ausfuehren, wenn der Benutzer angemeldet ist") umgestellt.
Kein gespeichertes Passwort noetig, da michi der dauerhaft angemeldete
Konsolen-User ist. Per Scheduler ausgeloest, Ergebnis 0x0 verifiziert
(SMB-Zugriff vorhanden, Spiegel frisch).
Doku korrigiert: README beschreibt jetzt Interactive als angewendete Loesung
(Password war nur die nicht genutzte Alternative). MASTER_TODO: Root-Cause
behoben, nur noch optionale Healthchecks-URL offen.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Diagnose 2026-06-21: Der Scheduled Task "KalliLab H Drive Nearline Pull"
lief als LogonType S4U (ohne gespeichertes Passwort) und hatte damit keine
Netzwerk-Anmeldeinformationen fuer den SMB-Share \192.168.178.58\backups.
Jeder geplante 05:30-Lauf brach still mit Exit 1 ab, ohne Report; der
Nearline-Spiegel war 2026-06-19 bis 2026-06-21 veraltet. Manuell nachgezogen,
Spiegel wieder frisch.
pull-critical-backups.ps1: explizites `exit 0` auf dem Erfolgspfad, damit der
letzte robocopy-Exitcode (1 = "Dateien kopiert") nicht als Prozess-Exit leakt
und der Scheduled Task ein wahrheitsgemaesses Ergebnis meldet.
README: Pflicht-Hinweis, dass der Task mit gespeichertem Passwort (nicht S4U)
laufen muss. MASTER_TODO: Root-Cause + verbleibender Operator-Schritt.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Schliesst den lokalen Code-Stand fuer zwei offene MASTER_TODO-Punkte ab.
monitoring: restliche Einzeldatei-Bind-Mounts (alertmanager, blackbox,
loki, promtail, alertmanager-ntfy-bridge) auf Directory-Mounts umgestellt,
analog zum Prometheus-Fix vom 2026-06-19. Vermeidet "Stale NFS file handle"
auf dem /mnt/user-FUSE-Share bei git/Komodo-Updates. grafana-provisioning
war bereits Directory-Mount. `docker compose config` gruen. Beim Deploy
--force-recreate noetig, da sich Mount-Zielpfade aendern.
backup: endpoint-agnostischer Dead-Man's-Switch (Healthchecks-kompatibel,
Cloud oder self-hosted) in pull-critical-backups.ps1 und pre-borg.sh.
Pings /start, Erfolg und /fail; No-Op ohne konfigurierte URL, bricht also
keinen Lauf. Ping-URLs sind Capability-URLs und bleiben als Secret
ausserhalb des Repos.
Doku: SECRETS_MAP, Nearline-README und MASTER_TODO nachgezogen.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
renovate
Micha