github-starred/pangolin
2
0
Fork 0
mirror of https://github.com/fosrl/pangolin.git synced 2026-05-21 09:21:15 -05:00
Code Issues 856 Packages Projects Releases 74 Wiki Activity

[GH-ISSUE #2550] Some requests hang indefinitely #15070

New Issue
Closed
opened 2026-05-16 03:22:32 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-05-16 03:22:32 -05:00
Owner
Copy Link

Originally created by @pf56 on GitHub (Feb 26, 2026).
Original GitHub issue: https://github.com/fosrl/pangolin/issues/2550

Describe the Bug

Some requests that are proxied through Pangolin/Newt seem to hang indefinitely for me.

Pangolin is running on a VPS (Hetzner), connected to Newt running at my home on a Kubernetes cluster.
Connectivity in general seems to be fine. The tunnel is connected and some requests are coming through.
Some other pages, however, load indefinitely.

As an example for a more common application that doesn't work:

  • accessing Sonarr on the URL exposed by Pangolin redirects to the login page, but hangs indefinitely
    Image

  • accessing the /ping endpoint of Sonarr returns OK (so the container running Sonarr must be accessible)
    Image

A similar issue happens with the Livekit React demo app (davidkornel/lk-react):
The initial index.html loads fine, but /static/js/bundle.js loads indefinitely.
Image

All of those pages work fine if I access them directly, either through a Kubernetes ingress or with a kubectl port-forward.

Now, here is where it's getting weirder:
I exec'd into the lk-react container (just a regular React application) and created some test files.
A test file of 758 bytes in size loads just fine. If I add another character (759 bytes), it hangs indefinitely.
(the files contain random text, created with base64 /dev/urandom | head -c [size] > test.txt)

758 bytes:
Image
759 bytes:
Image

To add even more confusion, at 1024 bytes the request starts working, but files at 1528 bytes break again (MTU related?).
lk-react also provides an index.html of 1607 bytes - that one loads just fine.
(I'd guess that the issue is somewhere on the network layer, and the HTML file is compressed to a size where it doesn't trigger the issue)

I collected tcpdump output showing a request to /static/js/bundle.js, both from the Pangolin host itself and the corresponding Newt container:
https://gist.github.com/pf56/c5458bc57040a021ea6c4c00a3a1537b

Environment

  • OS Type & Version:
    • Pangolin/Traefik/Gerbil: NixOS unstable (on nixpkgs commit 0182a36)
    • Newt: Kubernetes 1.34.2, Talos v1.11.5
  • Pangolin Version: 1.15.3
  • Gerbil Version: 1.3.0
  • Traefik Version: 3.6.7
  • Newt Version: tried 1.5.0 (default of Helm chart) and 1.10.0
  • Olm Version: (if applicable)

To Reproduce

I'm not really sure what's triggering the issue. Could be something network related on my end.
My best guess at how to reproduce:

  • deploy a docker container with davidkornel/lk-react
  • point a Pangolin instance at it, routed through Newt
  • see if the page comes up and /static/js/bundle.js can be loaded

If that doesn't reproduce the issue:

  • exec into the container
  • create files of varying sizes with base64 /dev/urandom | head -c [size] > test.txt
  • try to access the files (through the Pangolin url)

Expected Behavior

HTTP requests complete successfully, regardless of the size of the requested file

Originally created by @pf56 on GitHub (Feb 26, 2026). Original GitHub issue: https://github.com/fosrl/pangolin/issues/2550 ### Describe the Bug Some requests that are proxied through Pangolin/Newt seem to hang indefinitely for me. Pangolin is running on a VPS (Hetzner), connected to Newt running at my home on a Kubernetes cluster. Connectivity in general seems to be fine. The tunnel is connected and some requests are coming through. Some other pages, however, load indefinitely. As an example for a more common application that doesn't work: - accessing Sonarr on the URL exposed by Pangolin redirects to the login page, but hangs indefinitely <img width="1909" height="280" alt="Image" src="https://github.com/user-attachments/assets/3bb1968b-ea7d-4007-8ebb-944fcd0737fc" /> - accessing the `/ping` endpoint of Sonarr returns OK (so the container running Sonarr must be accessible) <img width="1907" height="254" alt="Image" src="https://github.com/user-attachments/assets/1069293e-c86b-4f6b-aaa8-b2ce2214a18a" /> A similar issue happens with the Livekit React demo app (`davidkornel/lk-react`): The initial `index.html` loads fine, but `/static/js/bundle.js` loads indefinitely. <img width="1906" height="331" alt="Image" src="https://github.com/user-attachments/assets/7d5725c5-43d5-4dd1-8309-24b1dc3e6aaf" /> All of those pages work fine if I access them directly, either through a Kubernetes ingress or with a kubectl port-forward. Now, here is where it's getting weirder: I exec'd into the `lk-react` container (just a regular React application) and created some test files. A test file of 758 bytes in size loads just fine. If I add another character (759 bytes), it hangs indefinitely. (the files contain random text, created with `base64 /dev/urandom | head -c [size] > test.txt`) 758 bytes: <img width="1911" height="290" alt="Image" src="https://github.com/user-attachments/assets/05e95087-6926-49fd-919d-77e9f0f4ed4f" /> 759 bytes: <img width="1907" height="288" alt="Image" src="https://github.com/user-attachments/assets/45015e74-763d-4d02-8d2c-1fe57cb1d49d" /> To add even more confusion, at 1024 bytes the request starts working, but files at 1528 bytes break again (MTU related?). `lk-react` also provides an `index.html` of 1607 bytes - that one loads just fine. (I'd guess that the issue is somewhere on the network layer, and the HTML file is compressed to a size where it doesn't trigger the issue) I collected tcpdump output showing a request to `/static/js/bundle.js`, both from the Pangolin host itself and the corresponding Newt container: https://gist.github.com/pf56/c5458bc57040a021ea6c4c00a3a1537b ### Environment - OS Type & Version: - Pangolin/Traefik/Gerbil: NixOS unstable (on nixpkgs commit `0182a36`) - Newt: Kubernetes 1.34.2, Talos v1.11.5 - Pangolin Version: 1.15.3 - Gerbil Version: 1.3.0 - Traefik Version: 3.6.7 - Newt Version: tried 1.5.0 (default of Helm chart) and 1.10.0 - Olm Version: (if applicable) ### To Reproduce I'm not really sure what's triggering the issue. Could be something network related on my end. My best guess at how to reproduce: - deploy a docker container with `davidkornel/lk-react` - point a Pangolin instance at it, routed through Newt - see if the page comes up and `/static/js/bundle.js` can be loaded If that doesn't reproduce the issue: - exec into the container - create files of varying sizes with `base64 /dev/urandom | head -c [size] > test.txt` - try to access the files (through the Pangolin url) ### Expected Behavior HTTP requests complete successfully, regardless of the size of the requested file
GiteaMirror commented 2026-05-16 03:22:34 -05:00
Author
Owner
Copy Link

@pf56 commented on GitHub (Feb 26, 2026):

Looks like it was an MTU issue on my network. Lowering the Newt MTU fixed it.

<!-- gh-comment-id:3966428329 --> @pf56 commented on GitHub (Feb 26, 2026): Looks like it was an MTU issue on my network. Lowering the Newt MTU fixed it.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
api
authentication
bug
config
dependencies
docker
documentation
enhancement
good first issue
help wanted
Improvement
Look Into
needs investigating
networking
new feature
non-critical bug
potential bug
pull-request
Mirrored from GitHub Pull Request
 question
reverse proxy
Security
stale
ui
wontfix
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/pangolin#15070
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?