Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

HTTP->HTTPS redirection assumes the request contains a host header

See original GitHub issue

Error description

Jupyterlab and proxy http to https work fine but getting this error. Log from /var/log/messeges:

Apr 28 03:20:33 jupyter jupyterhub[2015]: 03:20:33.871 [ConfigProxy] #033[31merror#033[39m: Uncaught Exception: Cannot read property 'split' of undefined
Apr 28 03:20:33 jupyter jupyterhub[2015]: 03:20:33.871 [ConfigProxy] #033[31merror#033[39m: TypeError: Cannot read property 'split' of undefined
Apr 28 03:20:33 jupyter jupyterhub[2015]:    at Server.<anonymous> (/usr/lib/node_modules/configurable-http-proxy/bin/configurable-http-proxy:382:35)
Apr 28 03:20:33 jupyter jupyterhub[2015]:    at Server.emit (events.js:198:13)
Apr 28 03:20:33 jupyter jupyterhub[2015]:    at parserOnIncoming (_http_server.js:691:12)
Apr 28 03:20:33 jupyter jupyterhub[2015]:    at HTTPParser.parserOnHeadersComplete (_http_common.js:111:17)

Systemctl:

$ systemctl status jupyterhub
● jupyterhub.service - Jupyterhub
   Loaded: loaded (/usr/lib/systemd/system/jupyterhub.service; enabled; vendor preset: disabled)
   Active: active (running) since Mon 2020-04-27 20:20:05 CEST; 12h ago
 Main PID: 2015 (jupyterhub)
    Tasks: 25 (limit: 23844)
   Memory: 104.5M
   CGroup: /system.slice/jupyterhub.service
           ├─2015 /usr/bin/python3.6 /usr/local/bin/jupyterhub -f /etc/jupyterhub/jupyterhub_config.py
           ├─2025 node /usr/bin/configurable-http-proxy --redirect-port 80 --ip jupyter.example.com --port 443 --api-ip localhost --api-port 8001 --error-target http://192.168.103.202:8443/h>
           └─2037 python3 /root/jupyterhub_services/cull_idle_servers.py --timeout=8640

Personal set up

OS: Linux 4.18.0-147.8.1.el8_1.x86_64 CentOS Linux release 8.1.1911 (Core)
Version: jupyterhub 1.1.0 configurable-http-proxy 4.2.1
Configuration: /etc/jupyterhub/jupyterhub_config.py

c.JupyterHub.cleanup_proxy = False
c.JupyterHub.cleanup_servers = False
c.ConfigurableHTTPProxy.command = ['configurable-http-proxy', '--redirect-port', '80']
c.ConfigurableHTTPProxy.api_url = 'http://localhost:8001'
c.JupyterHub.hub_ip = "192.168.103.202"
c.JupyterHub.hub_port = 8443
c.JupyterHub.bind_url = 'https://jupyter.example.com:443'
c.JupyterHub.ssl_cert = '/etc/jupyterhub/jupyter.crt'
c.JupyterHub.ssl_key = '/etc/jupyterhub/jupyter.key'
c.Spawner.default_url = '/lab'

Issue Analytics

State:
Created 3 years ago
Comments:6 (5 by maintainers)

Top GitHub Comments

1reaction

manicscommented, May 29, 2020

HTTP 1.1 requires a Host header, HTTP 1.0 doesn’t (though I didn’t think any browsers still used HTTP 1.0!). If you’ve only got one domain hosted on an IP you don’t need a host header since the certificate verification is done on the client side, and the client knows what address it connected to. If you’ve got multiple virtualhosts the server must support Server Name Indication and the client must send the host so the server knows which certificate to return, if it doesn’t the server normally returns a default certificate.

0reactions

belavicmcommented, Jun 3, 2020

I can’t reproduce the bug because my browser is not a problem (using HTTP 1.1). Maybe the problem is with one of the monitoring services (Zabbix, Wazuh, Puppet, etc.)?