Conda taking forever to install Jupyter lab

to get my kids back into the idea of school, I started doing some simple physics experiments with them. We’ve got some data and I wrote a Jupyter notebook to analyze. Then I went to install Jupyter lab on their computers… I am aware that IJulia has a built-in conda etc but I wanted to get them the “lab” version and have it be an independent install.

so jupyter lab says to install it via conda, and that means install conda… so I installed conda, and that took FOREVER, and then I install jupyter lab and that took FOREVER, and it’s still going.

And when I say FOREVER I mean like several HOURS. Is it normal that installing jupyterlab from scratch through conda is a 4 hour process that aborts a couple times before it completes? This is on Linux with a gigabit fiber connection. It’s not a network speed issue on my end.

Conda package operations can be excruciatingly slow sometimes. If you’re able to, I’d highly recommend GitHub - mamba-org/mamba: The Fast Cross-Platform Package Manager as a replacement.


The title of this thread is somewhat misleading as this issue has nothing to do with Julia.


Actually, once I got Jupyter lab installed, I went to ]add CSV,Plots,StatsPlots,DataFrames,IJulia and that ALSO took forever. Like maybe 2hrs… What the heck is going on?

Also, takes tens of minutes to load and shows long red bars on every site.

Note that I’m not trying to complain about Julia, I’m literally hoping someone will come along and say “hey none of this happens for me, and have you tried X” and then we’ll come up with a fix. It has to do with Julia because a bunch of people use Julia within Jupyter.

I’m beginning to wonder if this is a problem with connectivity to the package and content delivery sites? Perhaps an IPv6 routing issue or some other thing. I can’t ping on ipv4 or 6, I think it just doesn’t respond to pings. It’s quite confusing why any of this is taking nearly forever. I’m not sure which sites I should test for conda.

If the end goal is to use notebooks, then you might try the latest VS Code + Julia VS Code extension. It will use the built-in notebook capabilities for VS Code and send cell execution to the Julia extension. So, no need to install Conda, Jupyter, or IJulia.

As for the packages taking long, I’m not sure what the issue is. I haven’t had any problems recently. Maybe try the suggestion above, then let’s revisit the package issue if it persists (could have been a temporary issue)?


After experiencing extreme flakiness with the Pkg servers a while back, I (and probably many others) added this line to .julia/config/startup.jl

# Disable PKG server for now because it dies trying to download artifacts

Assuming you’re not on windows, this may help eliminate some of the more pathological behaviour.


If you blank out the PKG server env var how do you get packages? Does it fall back to GitHub or something?

1 Like

As I understand it, yes: JULIA_PKG_SERVER related documentation

I’ve got VSCodium, what do I do to get the latest stuff? If I open an .ipynb it just asks me what language it should use and doesn’t offer anything relevant.

AHA! it appears to have been a temporary bug in the VSCodium build, the latest version does open ipynb correctly: Jupyter notebooks not working · Issue #490 · VSCodium/vscodium · GitHub

However, it doesn’t know anything about any kernels. How do I install the kernels for VSCodium? Anyone know?

You need to install the latest Julia VS Code extension. it looks like VS Codium has its own re-packaged extension marketplace.

Thanks, I posted separately on this issue and it was mentioned that there’s a flag to check in the settings as well.

Julia + Jupyter notebooks in VSCodium?