Did the November 2024 AGOL Update Introduce an Overwrite Bug?
Update: I was having no luck solving my original problem, so I sucked it up and tore down the "corrupted" layer and its children sourcing my dashboard, loaded new data, tested that it would overwrite okay, generated hosted layers and secure services, and plugged those back in to my dashboard's maps and widgets. I called this a one-off overwrite() issue that was fixed with a new layer and moved on.
I ran my data pipeline code again (sourcing overwrite() from arcgis v1.3.5.post3) today and was displeased to see another, similar-but-new error.
Traceback (most recent call last):
...
File "~\agol_overwrite.py", line 240, in overwrite_agol_biweekly
flcollection.manager.overwrite(csvowriteversion_path)
File "~\venv\lib\site-packages\arcgis\features\managers.py", line 1785, in overwrite
published_item = related_data_item.publish(publish_parameters, overwrite=True)
File "~\venv\lib\site-packages\arcgis\gis\__init__.py", line 10384, in publish
elif not buildInitialCache and ret[0]['type'].lower() == 'image service':
KeyError: 'type'This error comes from a similar spot in the arcgis code to my original post, so I dug more.
The good news is that I have solved some of the mystery: errors arise from from trying to overwrite a base feature layer which has a hosted feature layer view downstream with a spatial filtering bounding box applied. Example below. I had a few points outside my area of interest, so I figured I could use this HFLV feature to focus my layer. I cannot remember when I set this feature, so I can't rule out that it wasn't the November 2024 that introduced changes here.
I loaded the same CSV used to supply my dashboard into AGOL and published it as fresh feature layers, created spatially filtered HFLVs downstream, and ran into my original issue. First time I was able to duplicated the original issue. When this bounding box was removed, I was able to overwrite in the GUI successfully. However, 1) I was no longer able to edit the HFLV's bounding box (see image below), and 2) the fix did not work for my main layer sourcing the public, production dashboard (naturally). Not being able to edit the bounding box again after it was tweaked once is consistent with my original issue and left me unable to test more then.
When I try to overwrite my primary feature layer now, I no longer get the same "failed" message in the GUI like the screenshot in the original post. I get a success message, but nothing happens; no features are added or deleted. The old record counts and table data remain. The data also don't record as being updated in the timestamp. It's like my overwrite command never happened without any notice of failure. No clue why this isn't matching my test FLs that I was able to fix with the spatial filtering rollback.
When I try with python, I continue to get the type publishing message above. There are several resources on addressing this out there (see here, here, here, here, to cite a few), but most of these are regarding the original .publish() method in the arcgis package, not that functionality embedded in .overwrite().
I'm back where I started with a feature layer that can't be overwritten sourcing a production dataset, only I know a tear-down, build-back solution won't solve the root problem. Any help here is welcome.
