We are diagnosing and fixing a problem with how user reports (old) are shimmed into user feedbacks (new).
The issue lies in event ingestion, as here we associate user reports with their events. It was assumed this only happened during user report ingestion, and the update_user_report task.
In order to fix this, we either have to:
Do the feedback shimming in ingestion here, or:
Don't link the event to the user feedback in this flow.
It would be preferred to do the feedback shimming directly in ingestion here. The overhead of this shouldn't be high, as all it really does is format the event correctly and emit a kafka message to the occurrence topic, but will double check the performance of this carefully.
In this first PR, behind a flag, we:
Change the UserReport query to only look at ones without an environment/group
Emit a metric when we update these records
The query needs to be changed as we want to get an accurate count of when we are actually updating records (update returns rows matched, not rows updated).
We are diagnosing and fixing a problem with how user reports (old) are shimmed into user feedbacks (new).
The issue lies in event ingestion, as here we associate user reports with their events. It was assumed this only happened during user report ingestion, and the update_user_report task.
In order to fix this, we either have to:
It would be preferred to do the feedback shimming directly in ingestion here. The overhead of this shouldn't be high, as all it really does is format the event correctly and emit a kafka message to the occurrence topic, but will double check the performance of this carefully.
In this first PR, behind a flag, we:
The query needs to be changed as we want to get an accurate count of when we are actually updating records (update returns rows matched, not rows updated).
Depends on https://github.com/getsentry/sentry/pull/70114
To Follow: https://github.com/getsentry/sentry/pull/70117