Notebook: Restarted computer, can no longer reach kernel

[Datamance]

After upgrading to Mac OS X 10.14.3 and restarting my computer, my notebook seems to be unable to reach the kernel - any cell I try to execute hangs with an asterisk forever.

I see this stacktrace my terminal:

```/Users/ricorodriguez/.local/share/virtualenvs/Math-Pm3XCPT4/lib/python3.7/site-packages/ipykernel_launcher.py:15: VisibleDeprecationWarning: zmq.eventloop.minitornado is deprecated in pyzmq 14.0 and will be removed.
Install tornado itself to use zmq with the tornado IOLoop.

from ipykernel import kernelapp as app
[I 00:32:26.337 NotebookApp] Adapting to protocol v5.1 for kernel 5617837b-16a3-4664-8186-ca371d51d698
/Users/ricorodriguez/.local/share/virtualenvs/Math-Pm3XCPT4/lib/python3.7/site-packages/notebook/base/zmqhandlers.py:284: RuntimeWarning: coroutine 'WebSocketHandler.get' was never awaited
super(AuthenticatedZMQStreamHandler, self).get(args, *kwargs)
RuntimeWarning: Enable tracemalloc to get the object allocation traceback
[W 00:32:27.519 NotebookApp] Replacing stale connection: 5617837b-16a3-4664-8186-ca371d51d698:ed21d4fb96ee4a0889e5ed93b19a627a
```

I've read a couple of things online about the websocket connection occasionally going bad. Do we know what's usually at the root of this?

[Datamance]

FWIW this notebook was serving locally and banging on all cylinders perfectly well last night. I literally just updated my computer, restarted and installed jupyterlab (uninstalled it to see if that was the problem, but no dice).

[Datamance]

After blowing out my pipenv environment and restarting, the kernel seems to be able to connect again - however I'm getting some strange behavior that wasn't happening before, namely plotly graphs that were working just fine before are not displaying at all.

[Datamance]

Nope, back to square one. After reloading some extensions I can see some web socket failures in the browser console...

[Log] Starting WebSockets: – "ws://localhost:8888/api/kernels/9bf6c96a-ff45-4210-a687-1bee11aa1133" (main.min.js, line 37334) [Error] WebSocket connection to 'ws://localhost:8888/api/kernels/9bf6c96a-ff45-4210-a687-1bee11aa1133/channels?session_id=9424407052da43c082db83bc522afbf0' failed: Unexpected response code: 200 [Log] Kernel: kernel_disconnected (9bf6c96a-ff45-4210-a687-1bee11aa1133) (main.min.js, line 36978) [Log] WebSocket connection failed: – "ws://localhost:8888/api/kernels/9bf6c96a-ff45-4210-a687-1bee11aa1133" – true (main.min.js, line 37419)

[wcass77]

I was having the same issue with installing jupyter on Pipenv. I would get it, with the same error messages, when I tried to install a new package and then it would go away when I synced to the old Pipfile.lock. I think the issue is that I had allow-prerelease turned on (needed for some other packages) and Pipenv was installing tornado 6.0 beta as a dependency. I added a line to my Pipfile to explicitly install tornado 5.1.1 or less (the latest stable release). This seems to have fixed the problem.

I hope that helps!

[alexis-mignon]

@wcass77 downgrading to tornado 5.1.1 solved the problem for me. Seems that there is a problem with tornado 6.0

[amacfie]

Commands if you installed with pip3:

sudo pip3 uninstall tornado
sudo pip3 install tornado==5.1.1

[vvk2001github]

sudo pip3 uninstall tornado
sudo pip3 install tornado==5.1.1

Thank!
This solved my problem.

[mikekaminsky]

What a nightmare this was to solve! Thank you @amacfie! Is there a way to pin the tornado version in this package somehow so that others don't go through this?

[DCtheTall]

Thanks @amacfie for the solution, I had been searching for an hour. Are the owners of this repo actively working on updating the code to support tornado 6, or is there a way to specify to pip that this version needs to be installed instead of the latest?

[junctionapps]

After a couple of hours toying with firewall rules as generally that's what is suggested with the forever 'connecting to kernel' message, I stumbled in here. Downgrading tornado works for me.

[pechyonkin]

Downgrading tornado from 6.0 to 5.1.1 solved the issue for me, too.

[mikeczabator]

sudo pip3 uninstall tornado
sudo pip3 install tornado==5.1.1

I had this same issue on a fresh install. I thought it was firewall issues or something else. Didn't realize it was due to tornado 6.0.

Thanks for the help.

[BramVanroy]

I can confirm that downgrading works. I had the same issue when using pipenv and installing from a Pipfile that didn't cause issues before.

[cosgriffc]

Just updated Jupyter this AM and ran into this. Downgrading tornado also fixed it for me.

[cxzbnm]

Commands if you installed with pip3:

sudo pip3 uninstall tornado
sudo pip3 install tornado==5.1.1

I was having this problem in Ubuntu 18.04 on virtual box. Wasted at least 6 hours trying to reinstall everything again and again. This solution was a lifesaver.

[audente]

Commands if you installed with pip3:

sudo pip3 uninstall tornado
sudo pip3 install tornado==5.1.1

Having the same problem on latest macOS Mojave. This solved it. Thanks.

[mdk73]

I got the same problem on Windows10, python 3.7, it could also be solved downgrading tornado. Thanks a lot.

[marcostrullato]

Worked for me - thanks!

[Kariap]

worked for me !! thanks man.... i was so confused using jupyter first time in my life and this error was making me pull my hair ! this solved it .

[nickagian]

I got the same problem on Windows10, python 3.7, it could also be solved downgrading tornado. Thanks a lot.

Exactly the same for me.

[bcoconni]

Same issue here. Fedora 28 64 bits and Python 3.6.8
I had to downgrade tornado to 5.1.1 for Jupyter to work again.

[MiroFurtado]

Worked for me as well! Changing the pip dependency should be a high priority issue

[hunalme]

Downgrading to 5.1.1 fixed my issue. Thanks.

[bjchiou]

Downgrading to 5.1.1 worked for me.

[valerio93105]

@wcass77 Downgrading to 5.1.1 worked for me. Thank you

[angelosalton]

Downgrading also worked for me.
Python 3.7.2, Windows 10 (x86-64)

[amacfie]

@mikekaminsky @DCtheTall A fix is on master https://github.com/jupyter/notebook/commit/9035cb1bdd2fe6c8195912f67d232ed77da0f6c0 but not released yet.

[leodambrosi]

Downgrading tornado from 6.0 to 5.1.1 solved the issue for me as well

[lmiguelvargasf]

notebook package has been updated to version 5.7.5 which fixes this issue due to Tornado 6 (see this tweet). Now, you can simply update the jupyter package.

For further details, you can also check this post on Stack Overflow.

[bcoconni]

notebook package has been updated to version 5.7.5 which fixes this issue due to Tornado 6 (see this tweet). Now, you can simply update the jupyter package.

Confirmed. Jupyter is working on my platform (Fedora 28/Python 3.6.8) with tornado 6.0 and notebook 5.7.5.

Thanks !

[nickliqian]

Downgrading to 5.0.2 worked for me, too.

[MarcusJones]

Ugh, just ran into this today, upgraded Jlab in my conda base env with conda update -c conda-forge jupyterlab, I suppose tornado 6 was already existing in my env from another package.

Maybe a quick patch assert tornado.version_info[0] == 5 with a nice message would help here in the meantime? Or is it always recommended to install jupyterlab into it's own env?