Merge development into production

[rbruhn]

Summary

• promote the current development branch to production

Included

• Remove duplicate transcriptions and enforce unique classification ids #482 Remove duplicate transcriptions and enforce unique classification ids (closes Remove duplicate Notes From Nature transcriptions and enforce unique classification ids #481)
  • app:update-queries pusher-transcription-duplicates keeps the earliest pusher_transcriptions document per classification_id, deletes the 401,231 extra documents, and rebuilds classification_id_1 as unique.
  • New migration: unique index on event_transcriptions (classification_id, event_id, team_id, user_id); EventTranscriptionService uses createOrFirst().

Deployment

Before merging, stop the Panoptes listener on production:

sudo supervisorctl stop biospex:panoptes-pusher

The deploy then runs the migration followed by the cleanup (update_queries_operation = pusher-transcription-duplicates); expect a few minutes on the shared MongoDB server.

After the deploy, start it again and confirm only one listener runs:

sudo supervisorctl start biospex:panoptes-pusher
sudo supervisorctl status | grep -i panoptes

Classifications missed while it was stopped are backfilled by the reconcile chain (ZooniversePusherJob). No events are running.

Follow-up: set update_queries_operation back to ''.

Validation

• development deployment completed on a fresh copy of production pusher_transcriptions with production's indexes:
  • 3,987,992 → 3,586,761 documents, one per classification_id (76,843 duplicated ids, 401,231 extra documents removed)
  • classification_id_1 is UNIQUE; the 711-copy classification 703618983 kept its earliest copy
  • event_transcriptions migration ran and its unique index exists