Login from internet not possible - cache issue - failure to fetch dynamically modules

Configuration Frontend
1

Hi,
short description of my landscape:
I have my own domain, and my HA is accessible via a subdomain.
I have a public DNS record at my domain provider that points to my Fritzbox.
The Fritzbox then uses a defined HTTPS port to forward the incoming traffic to my Home Assistant—or more precisely, to the Nginx proxy running on the Raspberry Pi, which also runs Home Assistant. Everything up to this point is via HTTPS. The Nginx proxy then forwards the traffic to the Home Assistant.From there, it’s only using HTTP because I haven’t managed to switch it to HTTPS for the homeassistant itself yet but thats a future topic.

It was working fine up to that point.
Since the last update of some components, including Home Assistant 2026.4.3, the page no longer loads from the internet.
The HomeAssistant login screen appears, so it seems that the path Subdomain → Fritzbox → Nginx → HomeAssistant is still working in general.
But then the login doesn’t work.

The internal access from my internal network directly to the homassistant http URL still works fine.

In the browser’s developer tools, I see the following error when accessing the site externally:
TypeError: Failed to fetch dynamically imported module: my_URL/frontend_latest/‘crypticNumber’.js
In the Nginx logs, I see the POST and GET requests and some http 499 errors while answering to /frontend_latest/.js . I searched for the http 499 and it seems this is especcially nginx error for “client closed connection before I could answer correctly”
And Home Assistant itself doesn’t log anything about this at all.

I came across a thread (After HA Update 2025.10.3; Error while loading page lovelace. · Issue #154735 · home-assistant/core · GitHub) from 2025 where someone had a similar issue with Cloudflare after an HA update; they resolved it:
“After deleting the CNAME entry for my Cloudflare tunnel, restarting the HA add-on for Cloudflare, and clearing the caches in Cloudflare, my browser, and AdGuard, everything is fine and access via the Cloudflare tunnel is working again.”

But as ther is no cloudflare in my architecture and I cannot delete any caches at my domain provider I don’t know how to solve my issue.
I already rebooted HA and nginx and deleted caches of my browser as well.

I cannot revert the updates as - as always - I waited too long to have a look at this issue and now my backups are already overwritten.

Hope someone has an idea.
Regards

Issue somewhere within reverse proxy, trusted_proxies, and DNS (#2)
2

It seems here the root cause is named:

->the cryptic number in the error message:
TypeError: Failed to fetch dynamically imported module: my_URL/frontend_latest/‘crypticNumber’.js
is the hashed filename of the component that is no longer available because I updated the HA. So it IS a cache issue but I still don’t know how to fix it because clearing the browser cache doesn’t work so it is another cache in my landscape and I neither know where not how to clear it :frowning:

3

Maybe that might be a current topic to resolve your issue?

Is AdGuard interfering?

4

Hi, I’m sry I’m new to this community. What do you mean with current topic? Should I (let) move this to another topic?

And to my issue:
I found an old backup from january and restored this - still the same issue. So it seems not to be dependent on my HA Update at all.

And I found Issue somewhere within reverse proxy, trusted_proxies, and DNS (#2)

Sadly there is also no reslolution :frowning:

No, no AdGuard involved.

5

Some other suggestions - as it seems not to be dependent to the HA version itself - perhaps there is a browser update that leads to this situation? I mean the chromium engine is used by several browsers and the “magic, it’s working again” in the other linked issue also point to a root cause and solution outside of HA environment.
Perhaps it also has something to do with the HSTS preload functionality activated in my nginx :thinking:

6

I still try to figure out the root cause. If I’m monitoring the requests in browser developer tools, a lot of requests are forwarded to my HA instance and also get a response (http 200, I see them also in nginx log) but others stay in status “pending” and either get a timout or forbidden after some time…
cannot find out why.

grafik
grafik1473×437 45.2 KB

And it’s not reproducable.
grafik

I can send the previous failed ‘mysubdomain_mydomain’.de/frontend_latest/31802.725b21e6a9fbfea6.js to the HA instance and get a valid respone and one minute later it’s not longer responding and then it is responding.
This leads to no conclusion… only confusing

7

I found my root cause: in my architecture it was a IPv4/IPv6 mismatch.
I just noticed that my homeassistant.local was responding randomly with ipv4 and ipv6 IP when I ping it in my local network.
And my nginx config was for homeassinstant.local (I used the DNS name not the ipv4 IP).
And in my configuration.yaml of HA in the trusted_proxies I only added the ipv4 address of the HA, not the ipv6.

Switching in nginx from DNS name to ipv4 for the homeassistant.local solved my issue.
I think I could also add the ipv6 to the trusted_proxies but I didn’t tried it yet.

ITS WORKING AGAIN :partying_face:

Edit: Still confused which of my components was updated and now used ipv6 instead but ipv4 but that doesn’t matter for me now as I solved it^^