Slow Undo Space Recycling Caused by Many Rollback Transactions

Symptom

The gs_async_rollback_xact_status view shows that there are a large number of transactions to be rolled back, and the number of transactions to be rolled back remains unchanged or keeps increasing.

select * from gs_async_rollback_xact_status();

Solution

Increase the number of asynchronous rollback threads in either of the following ways:

Method 1: Configure max_undo_workers in postgresql.conf and restart the node.

Method 2: Restart the instance using gs_guc reload -Z NODE-TYPE [-N NODE-NAME] [-I INSTANCE-NAME | -D DATADIR] -c max_undo_workers=100.