Slow S3 Write Speed of Storage Sink When Upstream Executes Massive 20000-Row Batch Deletes on a Single Table

[asddongmen]

Description:
When the upstream executes massive batch delete operations (20000 rows per batch) on a single table, the Storage Sink's write speed to S3 becomes abnormally slow.

Specific Phenomena:

1. ResolvedTs lag is normal;
2. CheckpointTs lag keeps increasing;
3. Sink memory quota is continuously full.

Key Data:
The Sink's write speed to S3 is only 400KB~2MB per second, which is completely insufficient to consume the upstream data writes.

Suspected Root Cause:
It is suspected that there is a bottleneck within the Storage Sink, which limits the overall processing speed.