I have been trying use Intake to index a new simulation I’ve run with MOM6-SIS2 (panan), but have been having some errors I’m not quire sure how to fix.
The code runs fine-ish (with several pynvml package FutureWarning warnings) in the building part:
import os
from dask.distributed import Client
from access_nri_intake.source.builders import Mom6Builder
client = Client(threads_per_worker = 1)
#building
path = “/path/to/my/simulations/folder”
builder = Mom6Builder(
path=path).build()
But when I try to save it…
builder.save(
name=“panan-01-zstar-2001rerun”,
description=“A branched re-run of Panan01 for the year 2001, with aditional heat budget and surface flux diagnostics”,
directory=‘/saving/path’)
update: it seems to be a problem with conda/analysis3-25.10, as it runs fine with conda/analysis3-25.11. conda Not sure if this is of interest for NRI people so I’ll leave this post here for a while before asking to delete it
Aidan
(Aidan Heerdegen, ACCESS-NRI Release Team Lead)
3
No need to delete it. Others might have the same issue and you have documented a fix. You can even mark you own reply as the solution to highlight this.
By the way if you want ACCESS-NRI attention for a topic it is best to tag it with help, then the triage team will see it, assess if it something that we can assist with, and if so find the person best placed to do so.
This was caused by some fairly fiddly packaging changes we made during September-November, where we had to migrate the intake-esm infrastructure in conda/analysis3 due to some budget cut related stuff going on in the US.
In the conda/analysis3-25.10 environment, loading data viaintake-esm should work fine, but trying to write datastores will cause issues.
For those who want to dig further, building intake-esm datastores actually relies on a separate package called ecgtools, and this was the source of this bug.