Altering table distribution can cause the data to be distributed incorrectly in GPDB

Products

• VMware Tanzu Greenplum

Issue/Introduction

When the GUC "gp_use_legacy_hashops=on" and a table is re-distributed with "ALTER TABLE table01 SET WITH (REORGANISE=true) DISTRIBUTED BY (col01);" then the data is distributed using the new distribution algorithm introduced in GPDB 6.x.

However, the meta data in the catalog shows that the table is distributed by the old distribution algorithm used prior to GPDB 6.x.

This can result in some queries not finding some tuples as they are on a different segment than expected.

Environment

Product Version: 6.3

Resolution

Workaround

Set "gp_use_legacy_hashops=off" and redistribute data for any table that is marked as using the old distribution algorithm.

Fix

A code fix is being developed and will be released in a future release.