Hi, We've been investigating some high db load in our HMS server (version 2.3.9 on Mysql 5.7 aurora 2.11.2). This seems to be due to sort indexing being created for queries on the COLUMNS_V2 table. After some digging we think we see the same thing as this ticket/PR tries to solve: https://issues.apache.org/jira/browse/HIVE-25293. The 'alter table.. cascade' exploding the number of rows in the COLUMNS_V2 tables. This is especially bad in tables with lots of partitions (10k+).
Anyone know of a good workaround or fix? So far we've been trying/thinking of: * adding indexes (doesn't help the returned set is still very large in some cases) * Potentially: recreating the Hive table with partitions to clean up the COLUMNS_V2. * Potentially: clean the COLUMNS_V2 (very risky) If anyone faced similar issues please let us know what you did. Patrick
