[GitHub] [hive] pvary commented on a change in pull request #2826: HIVE-21075 : Metastore: Drop partition performance downgrade with Postgres

[GitBox <gi...@apache.org>]

pvary commented on a change in pull request #2826:
URL: https://github.com/apache/hive/pull/2826#discussion_r758971384



##########
File path: standalone-metastore/metastore-server/src/main/java/org/apache/hadoop/hive/metastore/ObjectStore.java
##########
@@ -5275,38 +5275,67 @@ private void removeUnusedColumnDescriptor(MColumnDescriptor oldCD) {
       return;
     }
 
-    boolean success = false;
     Query query = null;
+    Query query2 = null;
+    boolean success = false;
+    LOG.debug("execute removeUnusedColumnDescriptor");
+    DatabaseProduct dbProduct = DatabaseProduct.determineDatabaseProduct(MetaStoreDirectSql.getProductName(pm), conf);
 
+    /**
+     * In order to workaround oracle not supporting limit statement caused performance issue, HIVE-9447 makes
+     * all the backend DB run select count(1) from SDS where SDS.CD_ID=? to check if the specific CD_ID is
+     * referenced in SDS table before drop a partition. This select count(1) statement does not scale well in
+     * Postgres, and there is no index for CD_ID column in SDS table.
+     * For a SDS table with with 1.5 million rows, select count(1) has average 700ms without index, while in
+     * 10-20ms with index. But the statement before
+     * HIVE-9447( SELECT * FROM "SDS" "A0" WHERE "A0"."CD_ID" = $1 limit 1) uses less than 10ms .
+     */
     try {
       openTransaction();
-      LOG.debug("execute removeUnusedColumnDescriptor");
-
-      query = pm.newQuery("select count(1) from " +
-        "org.apache.hadoop.hive.metastore.model.MStorageDescriptor where (this.cd == inCD)");
-      query.declareParameters("MColumnDescriptor inCD");
-      long count = ((Long)query.execute(oldCD)).longValue();
-
-      //if no other SD references this CD, we can throw it out.
-      if (count == 0) {
-        // First remove any constraints that may be associated with this CD
-        query = pm.newQuery(MConstraint.class, "parentColumn == inCD || childColumn == inCD");
+      // Fix performance regression for postgres caused by HIVE-9447
+      if (dbProduct.isPOSTGRES() || dbProduct.isMYSQL()) {
+        query = pm.newQuery(MStorageDescriptor.class, "this.cd == inCD");

Review comment:
       Is there a QueryWrapper which closes the query for us in a try-with-resources clause? 




-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: gitbox-unsubscribe@hive.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



---------------------------------------------------------------------
To unsubscribe, e-mail: gitbox-unsubscribe@hive.apache.org
For additional commands, e-mail: gitbox-help@hive.apache.org