Batch job error in Use Case - Crop

Dear all, I am going through the Notebook UC3 - Crop type feature engineering (rule-based), to see if we could include it in one of the upcoming ESA trainings.
In the section “Retrieve Timeseries”, I run the optional cell, and get this error with the batch job:



Thanks for advising us!

Post moved to OpenEO Platform category and bumped

Hi Amalia,
we’ll want to see what the actual error is. (This is normally printed, but your openEO client might be a version too old for that.)

Can you try something like this:
connection.job('job_id').logs()
Or else use ‘editor.openeo.cloud’, which should also show your jobs after logging in. Each job also has a button to show logs.

Thanks!
Jeroen

Hi Jeroen,

Ok so this is what I get after using connection.job(‘job_id’).logs()

Thanks,
Amalia

Apologies, I wasn’t clear, you want to replace ‘job_id’ with the actual id in your original screenshot. That’s the long one starting with vito-20…’

Oh of course ! :sweat_smile: (sorry I did it quickly yesterday and did not read well).
I re-run the cells, so the job_id has changed. Is this below making more sense? ( OpenEoApiError: [503] OidcProviderUnavailable: OIDC Provider is unavailable )

Hi Amalia,

that error from yesterday was actually caused by the authentication service from EGI being down.
However, thanks to your new job run, I was now able to find the error myself. This is what you should have seen:

The number of bands in the metadata 3/2 does not match the actual band count in the cubes (left/right): 3/3. You can fix this by explicitly specifying correct band labels.

Which is actually something we fixed in the random forest notebook but not in the rule based version.
The crucial thing is that the band names need to be set correctly after adding the ratio band to the Sentinel-1 cube, which can be done like this:

composite_s1 = s1.apply_dimension(dimension=“bands”,process=lambda x: array_modify(data=x, values=x.array_element(0)/x.array_element(1), index=0))
composite_s1 = composite_s1.rename_labels(“bands”, [“ratio”] + composite_s1.metadata.band_names)

Or you can try checking out the new version of the notebook. I didn’t run it to completion yet myself, so we may still encounter other small things that are not correct but did work at the time when we wrote that notebook.

Apologies for taking so long!