Upgrade to pangeo-notebook 2025.12.30

[jsignell]

Closes #52

Also:

• remove pins on pystac and geoviews now that newer versions are released
• remove icechunk and virtualizarr now that they are upstreamed
• remove stac_ipyleaflet Remove stac_ipyleaflet #51

[jsignell]

I think the failures are due to an issue with pytest>=9 and pytest-notebook (ref chrisjsewell/pytest-notebook#83)

[jsignell]

Hmmm. This passes locally.

[jsignell]

ah ha! I just learned that run_test.sh is not actually what gets run on CI. It is this: https://github.com/jupyterhub/repo2docker-action/blob/daa77f929833dcd836a711bc7837af945e9d27cc/create_docker_image.sh#L149-L168

[jsignell]

See jupyterhub/repo2docker-action#133

[jsignell]

Ok nice it seems like that change in repo2docker-action does the trick. I'm going to undo c0004ff now

[jsignell]

@yuvipanda suggested that we just use the fork to avoid pinning in the action itself.

[wildintellect]

@jsignell I can confirm that public.ecr.aws/nasa-veda/pangeo-notebook-veda-image:f20ed0ce5027 boots. Should we manually test some notebooks before merging?

[jsignell]

So far I have found two issues:

1. dask gateway never successfully scales the cluster. It seems like the scale request never really gets through to the cluster. So it's not even trying to get more workers. I haven't figured out yet whether this is jupyterhub-specific but it sure feels like it is.

reproducer:

from dask_gateway import Gateway

gateway = Gateway()
cluster = gateway.new_cluster()
cluster.scale(2)

Nothing ever happens. No workers come online and no workers seem to be trying to come online.

2. hvplot dynamic plots never update: moving the sliders, changing dropdowns have no impact. This one is almost certainly jupyterhub-specific

reproducer:

import hvplot.xarray  # noqa

air_ds = hvplot.sampledata.air_temperature("xarray")
air_ds.hvplot.image(data_aspect=1, frame_width=400, dynamic=True)

switch to a different time, image doesn't change.

[jsignell]

I just tried both of these in the existing pangeo image and they don't work there either. So while this isn't a great state of affairs, it doesn't seem to be worse.

[wildintellect]

@jsignell sounds like we should report upstream in a few places, but go ahead with this merge.

• Report broken docs in VEDA-docs
• Open a ticket about dask gateway with veda-jupyterhub/2i2c
• who should get the hvplot ticket?

[jsignell]

who should get the hvplot ticket?

Yeah I was trying to figure that out as well. I opened holoviz/holoviews#6787 in the meantime. Although I suspect this is hub/jupyter-myst specific.

[jsignell]

Ok! The jupyterhub team has resolved the dask-gateway issue and it looks like pinning down pyviz_comms has resolved the hvplot issue. So I think this one is good to merge and an improvement over the current default pangeo image.

[jsignell]

I had written it up in NASA-IMPACT/veda-jupyterhub#105 so the links are all in there.