N8n queue+workers on AKS: "Error connecting to n8n" popup

Questions
1

Describe the problem/error/question:

Error connecting to n8n
Could not connect to server. Refresh to try again.

This happens before entering any username or password. I checked the container logs, and nothing unusual appears.

I’m behind an Application Gateway Ingress Controller, and I can see some 401 Unauthorized responses in the access logs. Interestingly, when running n8n in main mode on the same cluster, the same 401s appear, but the popup does not show up.

If I log in with my user, everything works fine, workflows can be created, etc. The issue seems limited to pre-login frontend requests.

What is the error message (if any)?

Error connecting to n8n Could not connect to server. Refresh to try again.

Please share your workflow

Not Applicable

Share the output returned by the last node

Not Applicable

Information on your n8n setup

  • n8n version: 1.109.1
  • Database (default: SQLite): PostgreSQL (external)
  • n8n EXECUTIONS_PROCESS setting (default: own, main): queue + workers
  • Running n8n via (Docker, npm, n8n cloud, desktop app): Docker on AKS
  • Operating system: Linux (AKS nodes)
  • Redis: running in the same cluster
  • Application Gateway Ingress Controller: configured with HTTPS frontend, backend-protocol: Http
2

well looks like the error is likely your application gateway blockin pre-login API calls, check its rules to ensure endpoints like /healthz and /rest/settings are publicaly accessible for OPTIONS and GET requests.
i hope that helped.

3

hi Hasnain_Javed:

update:
The version with the popup issue is 1.109.1 or higher, since in version 1.108.2 the error popup does not appear. Everything indicates that the problem is the early call (before login) to the /rest/module-settings endpoint, which seems to be handled differently.

The error appears to be related to a change in the n8n v1.109.1 frontend, where it tries to load module settings (/rest/module-settings) before the user is authenticated. This results in a “401 Unauthorized” that the frontend interprets as a connection failure, instead of perhaps handling it as an unauthenticated state. In contrast, in version 1.108.2 that endpoint responds with “304 Not Modified.”

As additional information:
*** My Application Gateway is not blocking anything; it is currently configured only to log, not to block.
*** I am running n8n “Self-Hosted” in queue + worker mode on an AKS cluster.

4

yeah just downgrade it to v1.108.2 until a fix is published, or suppress the popup error in the UI code if you control it. The team already probably knows this regression.