Update to olcf-jupyter.rst about workflow HOT 5 CLOSED

Ok, interesting. I'll add a note to the page telling the user to include the conda.sh script (either through source or in their .bashrc file).

It might be that there is something else on your system that is set up to use port 8888? To be honest this is not something I've come across before, but good to hear you have a solution that works for you! FYI, the trick with using a different port for localhost is particularly useful if you wanted to connect to multiple JupterLab instances which are e.g. hosted on different remote machines 🙂

from workflow.

Hi Diana,

Thanks for your suggestions! So I just checked my .bashrc on summit and I have

. /sw/summit/python/3.7/anaconda3/5.3.0/etc/profile.d/conda.sh

in there already, which I assume I must have put there at some point in order to get conda to run correctly. The issue you linked to says that this is only a problem when conda activate my_env is executed in a script. I have only ever run this by running the commands directly in the terminal - did you find that you needed to include the conda.sh script in that case, or were you running things in a script?

I have run this in Safari on MacOS and Firefox and Chrome on Ubuntu and have always been able to just open the browser at localhost:8888, so I'm not sure the problem is with the address. If you're getting that error message, it suggests you have another Jupyter session open which is using the same port. You can fix this by opening the ssh session with another port on your local machine, e.g.

ssh -N -L localhost:8887:login1:8888 [email protected]

then you'd open up localhost:8887 in your browser.

from workflow.

Hi @harpolea ! Thanks for the super quick reply.

I think your option does the same as the "source" command I found, but it is a great idea to add it to .bashrc, instead of repeating it everytime.
I found the issue above, because there wasn't that line on my .bashrc, when I was running those commands on the Summit login node directly. To create the environment. Not in a script.

I think I understood the same from the issue that you wrote.
But even after I killed all processes (including terminal - shell - and safari) I still had that problem. Right after I saw the issue, I just opened another tab and added the tree part, without doing anything else, and it worked. So it might have been something else.

Thanks for the note on the change of the localhost! I will try that next time that I find this issue.

Cheers

from workflow.

Thanks @harpolea !
That is very useful indeed, thanks for the multiple localhosts suggestion :)

from workflow.

I think that now that there is a jupyterhub at OLCF, we can close this.

from workflow.