Hello everyone,

I set up my Unraid server about two months ago and I'm relatively new to this. For the past three days, I've been facing a persistent issue where I can no longer access my services via their public domains from my local network. External access (via cellular) works for some services.

I would be very grateful for your help!

Hardware & System

• Motherboard/CPU/RAM: MSI PRO B760M-P, Intel i3-12100, 32 GB DDR4 RAM

• Storage: 1 TB NVMe Cache, 3x 2TB Storage HDDs, 1x 2TB Parity HDD

• Unraid OS Version: 7.1.4

• Docker Version: 27.5.1

• Plugins: CA Mover Tuning, File Activity, Unraid Connect, Appdata Backup (not used so far), CA Auto Update Applications, Community Applications, Compose Manager, Dynamix Cache Directories, Dynamix File Integrity, Dynamix S3 Sleep, Dynamix System Information, Fix Common Problems, FolderView2, User Scripts

Network Setup

• Unraid Host IP: 192.168.178.105

• Gateway (Fritz!Box 7590): 192.168.178.1

• DNS (Pi-hole): 192.168.178.222

Services Overview

Container	Network/IP/Domain	Comment
Pi-hole (binhex)	br0, 192.168.178.222 pihole.local
[Formerly] Nginx Proxy Manager	192.168.178.105:8181 npm.local	(NPM) used this for most of the time
[Formerly] Caddy	not known anymore
[Currently] Traefik	bridge, 192.168.178.107:8080
Unraid-Cloudflared-Tunnel	bridge, n/a
Immich	bridge, 192.168.178.105:2283 im.mydomain.abc	compose stack with 4 containers
Nextcloud AIO	bridge, n/a (only mastercontainer) nc.mydomain.abc	9 containers
Jellyfin	bridge, 192.168.178.105:8096 jf.mydomain.abc
Home Assistant (VM)	192.168.178.110:8123 ha.mydomain.abc	Not a docker container but a VM (by default)
Paperless-ngx	bridge, 192.168.178.105:8000 no domain	additionally installed for this (advised by a tutorial): postgresql17, Redis, gotenberg, Apache-Tika-Server

Problem and previous solutions attempts

I've gotten everything to work perfectly. I could access all services via their IP addresses, their .local domains (with HTTPS via NPM), and their public *.mydomain.abc domains from both inside and outside my network. I've set up my subdomains for im/nc/jf/ha in NPM and got the Cloudflare tunnels working.

A few days ago, the last thing I installed was Paperless; no problems with that. Two days later, I added another Unraid share in Nextcloud (via external storage) which was when the problems started. But I can't imagine how this could be the cause of the problems, it may be a coincidence.

Suddenly, my Nextcloud froze. I restarted NPM and since then, it all went out of control. I can't even remember all the steps; I've chatted for hours with Gemini for days, trying to solve it.

Symptoms:

• Access to all *.mydomain.abc domains from within my LAN fails.

• Access via IP (e.g., http://192.168.178.105:8096) still works for most services.

• Access via cellular network (external) works for some services, but not consistently.

• Nextcloud is completely inaccessible as it relies on domain access.

• The .local domains also stopped working temporarily; by reinstalling and setting up NPM, I was able to fix this.

The browser errors are inconsistent, but the most common are:

• ERR_QUIC_PROTOCOL_ERROR (Chromium browsers)

• SSL_ERROR_UNRECOGNIZED_NAME_ALERT (Firefox)

• ERR_TOO_MANY_REDIRECTS

• 502 Bad Gateway or 504 Gateway Timeout (from Cloudflare, external access)

Troubleshooting Steps Already Taken

I have spent several days troubleshooting this extensively. The key steps that have already been tried without success are:

• Tried multiple browsers (Firefox, Chrome, Brave) on multiple devices to rule out client-side caching.

• Verified container-to-container communication using curl from within the proxy container.

• Tried multiple Docker network architectures (bridge, br0, custom networks) for the proxy and backend services.

• Complete reinstallation of reverse proxies (started with NPM, tested Caddy, currently on a fresh install of Traefik).

• Confirmed Pi-hole is correctly resolving internal requests to the proxy's IP using nslookup.

• Checked my DNS Records in Pi-hole.

• Fixed an initial issue where Pi-hole was also returning public IPv6 addresses.

• Reconfigured my Cloudflare tunnels.

• Deactivated QUIC/HTTP/3 in Cloudflare.

Gemini and ChatGPT have no ideas how to help anymore. Gemini officially gave up several times and told me to go to the Unraid forum. It said that the issue is a fundamental problem with the Unraid networking environment.

As I only installed Caddy and Traefik for testing if the problem was NPM, I think I would prefer to set up NPM again; I liked it more than Traefik.

I've attached my Unraid diagnostics file. And a txt file with the traefik.yml and config.yml and the current jellyfin labels for traefik (as example, currently only jellyfin is set up for testing in traefik). Gemini told me to send these.

Thank you very much for your help!

calypso-diagnostics-20250731-1151.zip traefik.yml, config.yml and jellyfin labels.txt

Bumping this to see if any of the networking experts have an idea what to investigate next. The problem still persists.

Meanwhile, I have reinstalled NPM and tested all variants of connection, see table below.

Thanks!