One of our important customers faced a problem with shadow_* tables: amount of records was so big that cron DB purge was blocked because of https://issues.redhat.com/browse/OSPRH-13035. We have used customized set of commands to clean shadow_* tables and reduced nova DB size from 50 GBs to few GBs.

At the same time, one problem wasn't addressed: numerous records were not removed from shadow_instance_extra table. It looks like this situation happened because problematic records had NULL deleted_at field.

Logs and DB dumps will be provided privately. It will be great to get an ack for manual shadow_instance_extra table cleanup using MySQL commands + fix this issue in upcoming RHOSP releases.