[BUG] - nodejs pinned in python-3.10 image
rokroskar opened this issue · comments
What docker image(s) are you using?
base-notebook
Host OS system
Ubuntu:22, Mac arm64
Host architecture
x86_64
What Docker command are you running?
docker run --rm -ti jupyter/base-notebook:python-3.10 cat /opt/conda/conda-meta/pinned
How to Reproduce the problem?
see above
Command output
No response
Expected behavior
Only see the python version pinned
Actual behavior
jupyter/base-notebook:python-3.10
image has the following pinned in /opt/conda/conda-meta/pinned
:
python 3.10.11
nodejs >=18.0
Anything else?
This seems to be the only image version where nodejs is pinned - if I repeat the same with latest
, python-3.8
, python-3.9
only the python version is pinned. This is causing problems trying to install additional packages that have different nodejs requirements.
So, the question is whether this pinned version is required? It seems to potentially be related to #1902
Latest Docker version
- I've updated my Docker version to the latest available, and the issue still persists
Hi @rokroskar.
Unfortunately, this is not the currently used python version in our images, so the image jupyter/base-notebook:python-3.10
will not get any further updates.
On the other hand, you can easily fix this issue on your own:
- You can inherit an image from
jupyter/base-notebook:python-3.10
and remove the line from/opt/conda/conda-meta/pinned
. - Or you can use the image which was pushed before #1902 was merged. The merge happened on 26th May, I've searched our images and it seems we also built our images May 25th, so you can use
jupyter/base-notebook:2023-05-25
: https://hub.docker.com/r/jupyter/base-notebook/tags?page=1&name=2023-05-25
Please, tell me if this helps.
~ > docker run -it --rm jupyter/base-notebook:2023-05-25 cat /opt/conda/conda-meta/pinned
python 3.10.11
Second options seems to work fine for me.
Hi @mathbunnyru thanks for the quick response - sure, I can do that. I was also more wondering if the fact that this pin is now gone is reverting the fix to #1901
I reverted the fix, because correct nodejs version was being installed by default so we no longer needed this fix.
#1914
Not sure how your image management works, but it seems that if this is causing unnecessary issues, an update to that tag could be pushed. I understand that's a potential can of worms though.
Well, we could in theory temporarily switch to building python-3.10 image.
But, in my opinion, it will create new problems for people using this tag and suddenly getting new Jupyter Notebook v7 and other updates they don't really expect.
And moreover, some newer tags might be overwritten as well, which might also affect some users.
It's by design we only support one set of images at a given time and try to move forward. Otherwise, tagging and building becomes really messy and time-consuming.
I see this to be a minor issue and easily fixable, so I don't think we will make an exception in this case, I hope you understand.
That being sad, I think I can close this issue, because I provided two solutions to the problem and we can't really fix old images (tags).
And thank you for creating an issue - if someone is gonna have the similar issue, they will be able to use the same solution(s).
Yep, that makes sense, thanks!