While working on
DM-24259, I performed automated Gen 2 ingest of an HSC dataset in the following order:
- Call obs.subaru.ingest.HscIngestTask on the raws, with the obs_subaru config overrides
- Call pipe.tasks.ingestCalibs.IngestCalibsTask on the calibs, with the obs_subaru config overrides
- Call lsst.pipe.tasks.ingestCuratedCalibs.IngestCuratedCalibsTask on the defects in obs_subaru_data, with the obs_subaru config overrides
This led to the following error on defects ingest:
Investigation revealed that the table had columns for date and filter, but the obs_subaru config for IngestCuratedCalibsTask expected the defects_visit table to only have a date column. Simon Krughoff speculates that the table was created during calib ingest (which is configured to have its *_visit tables be (date, filter)). However, if this is the case then it's not clear how ingestion passes when building obs_subaru; there are no instrument-specific config files for ingestion.
Given that defects should not be filter-dependent, this appears to be a bug in HSC ingestion. Investigate the issue to confirm the cause, and fix. One plausible option is to create the defects-related registry tables only when ingesting defects, not calibs.