Commit graph

11 commits

Author SHA1 Message Date
Viktor Barzin
aa58565ecc upgrade immich to v2.7.4 and increase rate limit burst
- Immich version: v2.7.3 → v2.7.4
- Immich rate limit: avg 200→500, burst 2000→5000 (both traefik and platform stacks)
2026-04-11 10:15:42 +01:00
Viktor Barzin
4d753a6486 fix(immich): improve thumbnail loading performance on iOS app
- Bump immich-server memory 1700Mi/2500Mi → 2000Mi/3500Mi to prevent OOM kills
- Disable anti-AI middleware chain for Immich (removes 3 unnecessary ForwardAuth
  hops per request — Immich content is behind auth, not crawlable)
- Double rate limit to 200 avg / 2000 burst for fast-scroll thumbnail requests
- Fix ImmichFrame image tag (1.7.4 → v1.0.32.0)
- Add PostgreSQL vector search prewarming and tuning (SSD storage type,
  init container for override conf, postStart pg_prewarm)
2026-04-08 08:08:53 +01:00
Viktor Barzin
f9e85964ce traefik: add middleware and platform traefik config updates 2026-04-06 11:57:52 +03:00
Viktor Barzin
0e3c0fb503 security: harden traefik auth flow — fix header spoofing, TLS leak, DERP rate-limit
- Auth-proxy fallback now sets ALL X-authentik-* headers (username, uid,
  email, name, groups) to prevent client-supplied header spoofing when
  Authentik is down. Previously only username was set, allowing a malicious
  client to inject fake X-authentik-groups.
- Catch-all IngressRoute restricted to *.viktorbarzin.me only. Non-matching
  domains no longer get the wildcard cert served (TLS info leak).
- Added rate-limit and CrowdSec middleware to catch-all IngressRoute.
- Added rate-limit middleware to Headscale DERP IngressRoute.
- Rotated auth-proxy basicAuth credentials (bcrypt cost 5 → 12, admin → emergency-admin).
- Created Authentik brute-force reputation policy (threshold -5, IP+username).
2026-04-05 20:01:06 +03:00
Viktor Barzin
878b556179 state(monitoring): update encrypted state 2026-03-29 01:04:11 +02:00
Viktor Barzin
c49e4561a3 consolidate MetalLB IPs: 5 → 1 (10.0.20.200)
Migrate all 11 LoadBalancer services to share 10.0.20.200:
- Update annotations: metallb.universe.tf → metallb.io
- Pin all services to 10.0.20.200 with allow-shared-ip: shared
- Standardize externalTrafficPolicy to Cluster (required for IP sharing)
- Remove redundant port 80 (roundcube) from mailserver LB
- Update CoreDNS forward: 10.0.20.204 → 10.0.20.200
- Update cloudflared tunnel target: 10.0.20.202 → 10.0.20.200

Services consolidated: coturn, headscale, kms, qbittorrent, shadowsocks,
torrserver, wireguard, mailserver, traefik, xray, technitium
2026-03-24 18:35:43 +02:00
Viktor Barzin
33037eba46 upgrade MetalLB v0.10.2 → v0.15.3 and update annotations
- Replace custom ViktorBarzin/metallb module with official Helm chart
- Migrate from ConfigMap-based config to CRD (IPAddressPool + L2Advertisement)
- Update Traefik LB annotations from metallb.universe.tf to metallb.io format
- Technitium DNS keeps stable IP 10.0.20.204 via MetalLB auto-assignment
- Headscale split DNS already configured to use 10.0.20.204
2026-03-24 17:24:05 +02:00
Viktor Barzin
1639910043 ingress latency: add histogram buckets, fix restarts, right-size memory
- Traefik: add fine-grained Prometheus histogram buckets (0.01-30s) for meaningful P50/P99
- Calibre: relax liveness probe (timeout 5→10s, threshold 3→6) to stop NFS-caused restarts
- Novelapp: increase memory 128Mi/256Mi → 640Mi/640Mi (confirmed OOMKilled, VPA upper 505Mi)
- Forgejo: increase memory 256Mi → 384Mi (at 80% of limit, VPA upper 311Mi)
- ActualBudget: add explicit resources to prevent silent LimitRange defaults
- Docs: update Nextcloud note from 4Gi → 8Gi limit (Apache spike history)
2026-03-23 10:52:43 +02:00
Viktor Barzin
21cfa8c072 bump memory limits for OOM-prone services
FreshRSS: 64Mi → 256Mi (171 restarts, VPA upper ~204Mi)
Actual Budget HTTP API: 128Mi → 384Mi (17 restarts, VPA upper ~297Mi)
n8n: 768Mi → 1Gi (18 restarts, VPA upper ~765Mi)
Dawarich: 768Mi → 896Mi (2 restarts, VPA upper ~628Mi)
Traefik: 384Mi → 768Mi (2 restarts, VPA upper ~584Mi)
2026-03-21 11:12:12 +00:00
Viktor Barzin
250a058627 feat(traefik): add custom error pages with tarampampam/error-pages
Deploy error-pages service to show themed error pages instead of raw
Traefik 502/503/504 responses. Adds catch-all IngressRoute (priority 1)
for 404 on unknown hosts. Only 5xx intercepted to avoid breaking JSON APIs.
2026-03-19 23:14:27 +00:00
Viktor Barzin
73511b1230 extract remaining 19 modules from platform, complete stack split [ci skip]
Phase 3: all 27 platform modules now run as independent stacks.
Platform reduced to empty shell (outputs only) for backward compat
with 72 app stacks that declare dependency "platform".
Fixed technitium cross-module dashboard reference by copying file.
Woodpecker pipeline applies all 27+1 stacks in parallel via loop.
All applied with zero destroys.
2026-03-17 21:42:16 +00:00