Problem:Kernel Restarting. The kernel appears to have died. It will restart automatically.

User 3543 | 3/13/2016, 4:28:30 PM

Hi there!

I am having problems when running jupyter-notebook. When I try to import sframe (import sframe), I am having the error message:

"Kernel Restarting. The kernel appears to have died. It will restart automatically."

It seems it is happening after upgrading the GraphLab-Create to version 1.8.4. I have seen this message in the "Dato Forum", that seems to be related:

However, after downgrading pyzmq to 14.6.0 version, the problem is still happening.

Are someone else having this problem?, do you have any solution?.

Thanks a lot for your help


My setup is: Description:

Ubuntu 14.04.4 LTS.

Python 2.7.6.

GraphLab-Create (1.8.4).

GraphLab-Create-License (1.8.1)

pyzmq 14.6.0


User 15 | 3/13/2016, 11:14:40 PM


Yes, to the best of our current knowledge, this is due to some ugly symbol collision that was inadvertently introduced in 1.8.4. We've noticed that downgrading pyzmq can be one way to mitigate the error, but it's only one possible way the error can present itself. We've also noticed the error go away if you downgrade GraphLab Create to 1.8.3, or use Anaconda Python instead of your system's Python. Feel free to try those out.

I apologize for the trouble this is causing. We hope to fix it soon.


User 3543 | 3/14/2016, 5:55:01 PM

Hi Evan,

I have downgraded GraphLab Create to versionn 1.8.3 and now it works.

Thanks for the help! Friman

User 3252 | 3/20/2016, 2:35:04 AM

I am using an aws image(coursera course) with version '1.8.3'. I am trying to follow the instructions in

My kernel dies soon after I deploy my service. I had to start over several times. How do I get a stable service? How do I clean up the previous deployments after the kernel restarted?

I think I am being charged by aws for the services deployed, which I am unable to track or terminate.

User 3252 | 3/20/2016, 3:08:51 AM

Update: I was working on AWS 'us-east datacenter, and found out that my predictive services were deployed on m3.xlarge instances in 'us-west2' data center. (Deployed multiple services because the kernel kept dying and it would not let me reuse the name after I start a new kernel). I manually terminated the instances. I also deleted the corresponding folders on us-east s3 bucket.

Hopefully there is no additional cleanup required.