How to solve “forbidden” error in jupyter notebook (conda tab)

[CorneeldH]

I installed anaconda and succeeded in making the conda-tab work thanks to the last suggestion in this entry: http://stackoverflow.com/a/41825208/6375668

However, when I try to create an environment or install an additional package I receive a "forbidden" error (see below). When I open anaconda-navigator I am perfectly able to create environments and install packages. In addition, I changed permissions to the anaconda-folders on my Mac and tried starting jupyter with sudo, but that didn't help.

What could be wrong? And how to fix this? Thanks!

[gnestor]

Are you seeing any notebook server errors in the terminal? What version of the notebook and nb-conda are you running?

Here's a similar issue on nb-conda: anaconda/nb_conda#44

[scaramouche-00]

Since updating my packages in anaconda (conda update --all) and installing doxygen, I have the same issue. Jupyter will throw the error
403 GET

[faizankshaikh]

Got the same error. Solved it by opening jupyter in another browser (was using firefox, then used chrome). I think its a "cache" related issue. Maybe someone can explore this futher?

[andreagrandi]

@faizankshaikh hi, I came to this page because I was having problems saving a notbook while using Firefox. If I use Chrome I don't have any problem, while if I use Firefox I keep getting the "Forbidden" error when I try to save a notebook. I've also tried to disable any extension but it didn't help.

[takluyver]

#3365 may be related.

[takluyver]

I've fixed one such issue in #3380, if people want to try that. But there may be more than one problem involved; the precise case I hit could only have come up after PR #2959, which was well after the initial report.

[scaramouche-00]

Hello, thanks for this! How can we access the patch? I have anaconda and I have tried updating jupyter (conda update jupyter), but it didn't work (still 403 GET).

[takluyver]

It will need to be merged and released before you can get it through Anaconda. For a small change like this, it's probably easiest to look at what I've changed, find the matching lines in your installation, and apply the change manually.

Otherwise, you could use git to get it:

git clone https://github.com/takluyver/notebook.git
cd notebook
git checkout i3365
pip install .

The last step needs some Javascript tools (see setting up a development environment).

[thomasaarholt]

Hammering F5 (refresh) seems to be a quickfix for this issue.

[pedrodcb]

Clear your cache then right click on and duplicate your browser tab containing the notebook. It will ask you for the Jupyter password again. Once you enter it successfully just go back to the original window and try saving, everything should work.

[vivianna-121]

Hammering F5 (refresh) seems to be a quickfix for this issue.

Exactly..

[mhmdmirzaei]

Probable reason:
One reason could be that your Jupyter is password protected and for some reason you need to re-enter the password.

Solution:
Open a new tab and try to connect to your Jupyter server in the new tab. If you are asked to re-enter the password, enter it and your problem should be solved. Go back to the original Jupyter tab and save the file. If you are not asked for the password, this solution doesn't work for you!

Problems with other solutions:
If you refresh or try to restart Jupyter, you might loose some of the work you haven't been able to save!

[yifanzhang13]

I've fixed using "sudo jupyter notebook": it requests the password and don't initialise as it does not understand the command, but reinstate the cache access (as it lost privileges for some reason). All works fine after that.

(tensorflow) VirtualBox:$ sudo jupyter notebook [sudo] password for user: sudo: jupyter: command not found (tensorflow) VirtualBox:$ jupyter notebook

I got the same problem today, this works for me.👍

[orwel1984]

Getting the same problem in MS Visual Studio Code.
Has anyone given that a try?

[EnisBerk]

I had the same issue on the MS Visual Studio Code,
I set the following in the config as suggested at vscode issue

c.NotebookApp.token = ''
c.NotebookApp.open_browser = False
c.NotebookApp.disable_check_xsrf = True

Also when you run a notebook, the server creates a tmp notebook in the directory jupyter server started, I did not have write access to that folder, that was the second issue.

[ghirardinicola]

Is the issue solved in newer versions or is it still there?

[stas00]

It's still there as of today.

Setting:

c.NotebookApp.token = ''

in ~/.jupyter/jupyter_notebook_config.py did the trick for me.

derived from @EnisBerk's #2396 (comment)

[OnlyBelter]

I use chrome always, and I get this occassionally. Go your home url (open another tab if necessary) and refresh that page. It will as ask for the authentication token, which can be found in the terminal where the notebook server was started. Once you are logged in (again), go back to your notebook tab and select the "kernel>reconnect" menu item.

I met this problem and I was unable to save the change. Just log in again from another tab, and return to the old one. The old one can save the change now. Thank @rsemenoff .

[lcd522]

Got the same error. Solved it by opening jupyter in another browser (was using firefox, then used chrome). I think its a "cache" related issue. Maybe someone can explore this futher?

Similar way. I Solved it by refreshing and re-enter my password.

[TommasoIacobellis]

I solved the problem by uninstalling Jupyter Notebook in the environment where I had the issue, from Anaconda console and then installing it again.

[lorenzznerol]

The following answer is only for those who work with a login.

Taking up another answer:

I also get this nasty error in chrome all of a sudden (it has been working for months now). If I use Edge, the problem is gone. Has anybody figured what it could be? I tried refreshing, deleting cache, cookies, reentered the token, ...

But still all requests error out with 403, except for the /login page:

[W 15:19:50.801 NotebookApp] 403 GET /api/sessions?_=1576160390722 (::1) 1.99ms referer=http://localhost:8888/tree
[W 15:19:50.802 NotebookApp] Forbidden
[W 15:19:50.802 NotebookApp] 403 GET /api/terminals?_=1576160390723 (::1) 1.99ms referer=http://localhost:8888/tree
[W 15:20:04.664 NotebookApp] Forbidden
[W 15:20:04.670 NotebookApp] 403 GET /api/sessions?_=1576160390724 (::1) 6.98ms referer=http://localhost:8888/tree
[W 15:20:04.675 NotebookApp] Forbidden
[W 15:20:04.676 NotebookApp] 403 GET /api/terminals?_=1576160390725 (::1) 2.00ms referer=http://localhost:8888/tree

Edit: I just realized using the local IP http://127.0.0.1:8888 instead of http://localhost:8888 works for some obscure reasons... At least for me

This, together with the answer just two cells before that at #2396 (comment) say that you just have to "re-log", that is also why you can enter the login page but no other page. If you work with a login, you need to refresh that login after a few days or weeks. I guess this has nothing to do with the browser, but with the server that you work on.

• The server might have a weekly logout time or a login limit.
• Or there might be a trigger to be logged out, for example, I had just run !pip install --upgrade torch and ran into:

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
fairseq 1.0.0a0+7818f61 requires cffi, which is not installed.
fairseq 1.0.0a0+7818f61 requires hydra-core<1.1, which is not installed.
fairseq 1.0.0a0+7818f61 requires omegaconf<2.1, which is not installed.
fairseq 1.0.0a0+7818f61 requires sacrebleu>=1.4.12, which is not installed.
torchvision 0.12.0 requires torch==1.11.0, but you have torch 2.1.2 which is incompatible.
torchaudio 0.9.0 requires torch==1.9.0, but you have torch 2.1.2 which is incompatible.
Successfully installed mpmath-1.3.0 nvidia-cublas-cu12-12.1.3.1 nvidia-cuda-cupti-cu12-12.1.105 nvidia-cuda-nvrtc-cu12-12.1.105 nvidia-cuda-runtime-cu12-12.1.105 nvidia-cudnn-cu12-8.9.2.26 nvidia-cufft-cu12-11.0.2.54 nvidia-curand-cu12-10.3.2.106 nvidia-cusolver-cu12-11.4.5.107 nvidia-cusparse-cu12-12.1.0.106 nvidia-nccl-cu12-2.18.1 nvidia-nvjitlink-cu12-12.3.101 nvidia-nvtx-cu12-12.1.105 sympy-1.12 torch-2.1.2 triton-2.1.0

This answer on Stack Overflow at Cannot connect to Jupyter Notebook also says that it could be a pip installation that leads to this:

Using pip to install some packages resulted in confusing the jupyter installation.

No matter whether it is a server timeout or triggered by some installation, there is an easy fix for this, at least if you were just logged out:

• open another tab of the same browser session,
• log in on your Jupyter Notebook server there,
• you can close that tab,
• go back to the tab that is forbidden,
• click on "forbidden",
• and it will give way to work on the page again.

I found this out by following the remarks here to press F5 and the login screen came up, and it seems as if you do not even lose any unsaved work. But the smartest way that makes 100% sure that you are not thrown back to any autosaved time before is to log in at a new tab of the same session.