Determining the Per-Node Storage of a Segmented Table

Jim Knicely authored this blog. You can join a few system tables together to find the disk space used and the row counts by node for segmented tables: Example: dbadmin=> SELECT ps.anchor_table_schema, dbadmin-> ps.anchor_table_name, dbadmin-> ps.node_name, dbadmin-> ROUND(SUM(ps.ros_used_bytes)/1024/1024/1024, 5)::NUMERIC(25,5) AS GB, dbadmin-> SUM(ps.ros_row_count) AS Rows dbadmin-> FROM v_catalog.projections p dbadmin-> JOIN v_monitor.projection_storage ps dbadmin-> USING (projection_id) dbadmin-> WHERE p.is_segmented dbadmin-> AND ps.anchor_table_schema || '.' || ps.anchor_table_name = 'public' || '.' || 'big_table' dbadmin-> GROUP BY 1, 2, 3 dbadmin-> ORDER BY 1, 2, 3; anchor_table_schema | anchor_table_name | node_name | GB | Rows ---------------------+-------------------+--------------------+---------+----------- public | big_table | v_test_db_node0001 | 1.70385 | 810901653 public | big_table | v_test_db_node0002 | 1.70385 | 810919791 public | big_table | v_test_db_node0003 | 1.70395 | 810968034 (3 rows) Skew in the data distribution across the nodes is an indicator of a poor choice of the table’s segmentation key. Have fun!