2

Im trying to install an allegroai trains-server on a k8s cluster.

I tried the following 3 methods

I followed the linux installation to the letter, and in the k8s installations used the following command to access the exposed port of the trains webserver kubectl port-forward -n trains svc/webserver-service 9999:80

In all three cases i manage to get to the server, but it looks empty and most operations fail.Here is a screen shot of what the webserver looks like.

I tried doing all 3 multiple times from scratch and even rebuilt my k8s cluster but nothing works.

Does anyone know how to solve this?

DalyaG
  • 2,979
  • 2
  • 16
  • 19
Dean Light
  • 21
  • 1
  • 1
    Try asking at [unix.se] or [sf]. – 001 Nov 03 '20 at 16:30
  • 1
    The trains repo says to ask here with the trains tag. – Dean Light Nov 03 '20 at 16:53
  • Turns out this was a token caching problem. When you log in to a new server but have an auth token of the old server, you get a broken interface instead of it bringing you to a new login screen. Cleaning my cookies solved the problem. – Dean Light Nov 04 '20 at 18:46

1 Answers1

0

Disclaimer - I'm part of the allegro.ai Trains team

The current implementation of trains-server supports two types of deployments

  1. Single IP with fixed set of ports:
  • Port 8080 - Web service,
  • Post 8008 - API
  • Port 8081 - Generic http file serving
  1. sub-domain deployment:

There is also a helm chart deployment that will deploy the trains-server on a single node exposing ports 30080/30008/30081 You can find the details here

Martin.B
  • 599
  • 3
  • 9