Hi all, I’m having recent trouble trying to read netCDF files in Python using xarray on my ARE gadi session (Jupyter Notebook). I think it’s an environment/package issue. The line that it struggles with is: ds = xr.open_mfdataset(yearly_list, engine="netcdf4", chunks={"time": 1200, "realization": …

Hi Charles, Thanks for your reply! Unfortunately it doesn’t seem to change the issue if I load the openmpi directly. But I found a fix - if I use analysis3-25.12 it seems to work now! It may be an issue with older versions or analsysi3-edge? I had to explicity change it to this version Thanks so m…

Hey Joel, It looks like for some reason the _edge environment is getting dragged in there. I’m not sure what’s going on with the 25.03 environment at this point, can you reproduce the error in the analysis3-stable environment (just module load conda/analysis3 should get you the one I’m after) and t…

Hi Charles, Thanks so much for the reply. I’ve loaded the stable analysis3 environment (analysis3-25.11) and it has the same issue unfortunately. The error looks slightly longer but looks to be the same problem. Cheers, Joel -----------------------------------------------------------------------…

Yeah, that’s very weird, and definitely not what I was hoping to see. Please could you post a screencap of your ARE session setup for me? I’m hoping maybe there’ll be some clues in there…

FWIW, I’m able to open datasets using xarray without any issues, so I’m hoping this is a config issue sneaking in somehow and not an environment problem. I’ve uploaded a screenshot of my config - hopefully you can spot someting in there which looks meaningfully different to yours? (Ignoring all the…

Hi Charles, Thanks for verifying that it works on your end - makes the issue even more odd. I can’t see anything that’s overly different except for you loading openmpi. Here is my setup: [Screenshot 2025-12-19 at 5.13.36 pm] [Screenshot 2025-12-19 at 5.14.03 pm] [Screenshot 2025-12-19 at 5.13…

The error you’re getting is related to xarray handling parallel worker communication (dask stuff), which makes me wonder if explicitly loading openmpi is stopping me seeing the issue. Can you try loading openmpi like I have and let me know if that fixes it?

Weird - the edge environments are more likely to have this sort of issue, but I wouldn’t expect to see it in the regular analysis3-25.11 environment. Glad you’ve found a fix - I’ll see if I can reproduce the error when I get the chance.

Reading netCDF in Python: Issue within analysis3?

Technical

CharlesTurner (Charles Turner) Split this topic 26 February 2026 01:32 12

A post was split to a new topic: Analysis3-edge, netCDF4 & PyTorch CUDA

Topic		Replies	Views
'NetCDF: Not a valid ID' errors Technical	26	3457	29 July 2024
Analysis3-edge, netCDF4 & PyTorch CUDA Technical python , netcdf	16	117	17 March 2026
Issues with new conda/analysis3 importing netCDF4 Infrastructure help , conda	4	53	13 April 2026
Cannot print to file - possible Dask issue in xp65 General help	18	100	11 August 2025
Unable to load netCDF4 module Technical python , help , jupyterlab	7	172	2 December 2025

Reading netCDF in Python: Issue within analysis3?

Related topics