initialization-actions: Jupyter creating or open notebooks causes 500 Internal Server Error

The initialization action for installing Jupyter and connecting to the web interface is successful. However opening an existing notebook or creating a new notebook results in a 500 : Internal Server Error. Screen Shot 2019-03-13 at 10 07 59 AM

About this issue

Original URL
State: closed
Created 5 years ago
Reactions: 3
Comments: 16 (5 by maintainers)

Most upvoted comments

@michelle-unity with Dataproc 1.4 you need to use components instead of init actions (see https://github.com/GoogleCloudPlatform/dataproc-initialization-actions/issues/480#issuecomment-472516743):

--optional-components=ANACONDA,JUPYTER \

medb on Mar 13, 2019

For Python 3 you need to use Dataproc 1.4 (--image-version=preview) with Jupyter Component.

medb on Mar 13, 2019

I’m using --image-version=1.2 and also got the same error this morning. When I checked /var/log/jupyter.log on the master node, I saw the following error:

AttributeError: module 'tornado.web' has no attribute 'asynchronous'

I then followed a recommendation to downgrade to conda install tornado=5.1.1, and restarted Jupyter /usr/local/bin/launch_jupyter.sh

This fixed the issue in my currently running cluster.

Unfortunately, the original error caused my notebook to be overwritten in the bucket with

{
 "cells": [],
 "metadata": {},
 "nbformat": 4,
 "nbformat_minor": 2
}

Good that I created a backup beforehand, and now have bucket versioning turned on to prevent loosing my work in the future.

sukhinin541 on Mar 14, 2019

Tried the init action with image-version=1.2, the same issue still persist.

MaxAVF on Mar 13, 2019