Navigating to my.domain.tld/collections/uuid works but bookmarked my.domain.tld/collections/uuid shows 502 Bad Gateway

[Jacob-Lasky]

If I navigate from my.domain.tld/collections --> my.domain.tld/collections/uuid then it works fine:

---

But if I navigate paste the URL and try to navigate directly to it (i.e. if I share it) then I'm met with the 502:

GET
https:/my.domain.tld/collections/uuid
[HTTP/2 502  249ms]

---

Additionally, I cannot download attachments as I receive a 500 error.

I've generally struggled to get the Docker Compose working in UnRAID with a public URL via NginxProxyManager. I've read some of the support docs and discussion but I'm sure that's where my issue is. I have two domains for AdventureLog, one pointing to mymedia.domain.tld and one pointing to my.domain.tld Here is the compose:

networks:
  adventurelog:
    driver: bridge
    
services:
  db:
    image: postgis/postgis:15-3.3
    container_name: adventurelog-db
    restart: unless-stopped
    environment:
      POSTGRES_DB: [REDACTED]
      POSTGRES_USER: [REDACTED]
      POSTGRES_PASSWORD: [REDACTED]
    ports:
      - "1925:5432"
    volumes:
      - /mnt/user/appdata/adventurelog/db:/var/lib/postgresql/data/
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U adventure -d database || exit 1"]
      interval: 10s
      timeout: 5s
      retries: 10
    networks:
      - adventurelog

  server:
    image: ghcr.io/seanmorley15/adventurelog-backend:latest
    container_name: adventurelog-backend
    restart: unless-stopped
    environment:
      # --- Postgres ---
      PGHOST: [REDACTED]
      PGDATABASE: [REDACTED]
      PGUSER: [REDACTED]
      PGPASSWORD: [REDACTED]

      # --- Django / App ---
      SECRET_KEY: [REDACTED]
      DEBUG: "False"

      # Public URLs
      PUBLIC_URL: "https://mymedia.domain.tld"
      FRONTEND_URL: "https://my.domain.tld"

      # CSRF
      CSRF_TRUSTED_ORIGINS: "https://my.domain.tld,https://mymedia.domain.tld"

      # initial admin user
      DJANGO_ADMIN_USERNAME: [REDACTED]
      DJANGO_ADMIN_PASSWORD: [REDACTED]
      DJANGO_ADMIN_EMAIL: [REDACTED]
      
      # Behind NPM/HTTPS, tell Django to trust X-Forwarded-Proto
      USE_X_FORWARDED_HOST: "true"
      SECURE_PROXY_SSL_HEADER: "HTTP_X_FORWARDED_PROTO,https"
      SESSION_COOKIE_SECURE: "true"
      CSRF_COOKIE_SECURE: "true"
    ports:
      - "1924:8000"
    depends_on:
      db:
        condition: service_healthy
    volumes:
      - /mnt/user/appdata/adventurelog/media:/code/media/
    networks:
      - adventurelog

  web:
    image: ghcr.io/seanmorley15/adventurelog-frontend:latest
    container_name: adventurelog-frontend
    restart: unless-stopped
    environment:
      PUBLIC_SERVER_URL: "https://mymedia.domain.tld"
      ORIGIN: "https://my.domain.tld"

      # If the image honors this; otherwise manage body size at the proxy
      BODY_SIZE_LIMIT: "Infinity"
    ports:
      - "1923:3000"
    depends_on:
      - server
    networks:
      - adventurelog

[domhaas]

Same issue here - running adventurelog behind a nginx-reverse-proxy.
e.g. "/collections" works

other URLS are not working, like:
"/collections/49b3fe25-6245-4dd5-ae5e-..."
"/locations"

-> 502 Bad Gateway (openresty)

But only if I do a F5-Refresh or using a bookmark - navigating inside the web-app works without issues.
No Errors occurring in the logs.

---

I think there are properties missing on these affected pages.

[Epithumia]

The issue is caused by a mix of Svelte and Nginx : The SvelteKit frontend server responds with larger headers on refresh, which the default nginx/NPM configuration cannot handle. This is shown when looking at the error log:
2025/11/23 23:06:47 [error] 8126#8126: *4407020 upstream sent too big header while reading response header from upstream, client: X.X.X.X, server: my.server.tld, request: "GET /locations HTTP/2.0", upstream: "http://Y.Y.Y.Y:3000/locations", host: "my.server.tld"

Solution is simple when you know what the true issue is, you just need to add the following three lines to the advanced tab of your reverse host:

proxy_busy_buffers_size   512k;
proxy_buffers   4 512k;
proxy_buffer_size   256k;

[Jacob-Lasky]

Oh brilliant, that's exactly what I needed. That fixed it right up, thank you!

…

On Sun, Nov 23, 2025, 5:16 PM Rafael Lopez ***@***.***> wrote: The issue is caused by a mix of Svelte and Nginx : The SvelteKit frontend server responds with larger headers on refresh, which the default nginx/NPM configuration cannot handle. This is shown when looking at the error log: 2025/11/23 23:06:47 [error] 8126#8126: *4407020 upstream sent too big header while reading response header from upstream, client: X.X.X.X, server: my.server.tld, request: "GET /locations HTTP/2.0", upstream: " http://Y.Y.Y.Y:3000/locations", host: "my.server.tld" Solution is simple when you know what the true issue is, you just need to add the following three lines to the advanced tab of your reverse host: proxy_busy_buffers_size 512k; proxy_buffers 4 512k; proxy_buffer_size 256k; — Reply to this email directly, view it on GitHub <#828 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABTZYXIABICYBGQAM33LFLT36IW4HAVCNFSM6AAAAACFLUIU26VHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTKMBVGUZDGMA> . You are receiving this because you authored the thread.Message ID: ***@***.*** com>