GPDB backup Performance Degradation after upgrading from gpbackup 1.33.0 to gpbackup 1.33.2

Products

• VMware Tanzu Data
• VMware Tanzu Greenplum

Issue/Introduction

When using gpbackup version 1.33.2 with the S3 plugin, backup and restore operations take significantly longer compared to version 1.33.0. This performance degradation compounds and becomes particularly noticeable when processing databases with a large number of tables.

Environment

 

Application: gpbackup v1.33.2

Storage Target: AWS S3 (via the gpbackup S3 plugin)

Condition: Schemas/Databases containing high table counts

 

Cause

In gpbackup version 1.33.2, the underlying S3 plugin was upgraded from AWS SDK version 1 to AWS SDK version 2.

When logging performance metrics, the software retrieves individual file sizes from S3. Under AWS SDK version 2, retrieving a file's size requires an extra round-trip network request to validate the file's checksum. While this additional latency is negligible on a single file, the per-table overhead accumulates rapidly during large-scale backups and restores, leading to severe bottlenecks.

Resolution

This issue will be resolved in an upcoming release of gpbackup 1.33.3.

Because file sizes are collected exclusively for internal performance logging, engineering is removing the redundant S3 network check and replacing it with a local tracking mechanism. This eliminates the network overhead without impacting logging data.

Additional Information

If the backup performance degradation is impacting critical operations and you cannot wait for the upcoming patch, consider the following temporary mitigation:

• Downgrade: Temporarily revert to gpbackup version 1.33.0, which utilizes AWS SDK version 1 and does not incur the extra checksum validation overhead.