You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I just investigated some cases of failed jobs, which wrote a store, but failed the test stage. I propose to also log these stores, but in a different BigQuery table.
Some background:
iid: CMIP6.ScenarioMIP.NIMS-KMA.UKESM1-0-LL.ssp585.r14i1p1f2.day.sfcWind.gn.v20210428 submitting gh workflow Dataflow job
Zarr Store: gs://leap-persistent-ro/data-library/cmip6-testing/a618127503-5957190341-2/CMIP6.ScenarioMIP.NIMS-KMA.UKESM1-0-LL.ssp585.r14i1p1f2.day.sfcWind.gn.v20210428.zarr
Failed due to:
RuntimeError: AssertionError [while running 'Create|OpenURLWithFSSpec|OpenWithXarray|Preprocessor|StoreToZarr|TestDataset|LogToBigQuery/TestDataset/Testing - Time Dimension-ptransform-53']
indicating that the test stage failed.
Investigating the store a bit more, it turns out that the time has gaps.
Now I think that fixing this sort of thing is beyond our mission here, but some researchers might want to investigate this more, and we should not keep them from it. I think this goes back to the idea @naomi-henderson established of a QC/non-QC catalog.
This would also help preventing to run these again and again, as they will fail every time.
The text was updated successfully, but these errors were encountered:
I just investigated some cases of failed jobs, which wrote a store, but failed the test stage. I propose to also log these stores, but in a different BigQuery table.
Some background:
iid:
CMIP6.ScenarioMIP.NIMS-KMA.UKESM1-0-LL.ssp585.r14i1p1f2.day.sfcWind.gn.v20210428
submitting gh workflow
Dataflow job
Zarr Store:
gs://leap-persistent-ro/data-library/cmip6-testing/a618127503-5957190341-2/CMIP6.ScenarioMIP.NIMS-KMA.UKESM1-0-LL.ssp585.r14i1p1f2.day.sfcWind.gn.v20210428.zarr
Failed due to:
indicating that the test stage failed.
Investigating the store a bit more, it turns out that the time has gaps.
Now I think that fixing this sort of thing is beyond our mission here, but some researchers might want to investigate this more, and we should not keep them from it. I think this goes back to the idea @naomi-henderson established of a QC/non-QC catalog.
This would also help preventing to run these again and again, as they will fail every time.
The text was updated successfully, but these errors were encountered: