[HIVE-10296] Cast exception observed when hive runs a multi join query on metastore (postgres), since postgres pushes the filter into the join, and ignores the condition before applying cast

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.13.1
• Fix Version/s: 3.2.0, 4.0.0-alpha-1
• Labels: pull-request-available
• Assignee: Karthik Manamcheri
• Reporter: Yash Datta
• Created: 10/Apr/15 05:22
• Updated: 17/Nov/22 08:54
• Resolved: 01/Oct/18 15:41

Description

Try to drop a partition from hive:

ALTER TABLE f__edr_bin_source__900_sub_id DROP IF EXISTS PARTITION ( exporttimestamp=1427824800, timestamp=1427824800)

This triggers a query on the metastore like this :

"select "PARTITIONS"."PART_ID" from "PARTITIONS" inner join "TBLS" on "PARTITIONS"."TBL_ID" = "TBLS"."TBL_ID" and "TBLS"."TBL_NAME" = ? inner join "DBS" on "TBLS"."DB_ID" = "DBS"."DB_ID" and "DBS"."NAME" = ? inner join "PARTITION_KEY_VALS" "FILTER0" on "FILTER0"."PART_ID" = "PARTITIONS"."PART_ID" and "FILTER0"."INTEGER_IDX" = 0 inner join "PARTITION_KEY_VALS" "FILTER1" on "FILTER1"."PART_ID" = "PARTITIONS"."PART_ID" and "FILTER1"."INTEGER_IDX" = 1 where ( (((case when "TBLS"."TBL_NAME" = ? and "DBS"."NAME" = ? then cast("FILTER0"."PART_KEY_VAL" as decimal(21,0)) else null end) = ?) and ((case when "TBLS"."TBL_NAME" = ? and "DBS"."NAME" = ? then cast("FILTER1"."PART_KEY_VAL" as decimal(21,0)) else null end) = ?)) )"

In some cases, when the internal tables in postgres (metastore) have some amount of data, the query plan pushes the condition down into the join.
Now because of DERBY-6358 , case when clause is used before the cast, but in this case , cast is evaluated before condition being evaluated. So in case we have different tables partitioned on string and integer columns, cast exception is observed!

15/04/06 08:41:20 ERROR metastore.ObjectStore: Direct SQL failed, falling back to ORM
javax.jdo.JDODataStoreException: Error executing SQL query "select "PARTITIONS"."PART_ID" from "PARTITIONS" inner join "TBLS" on "PARTITIONS"."TBL_ID" = "TBLS"."TBL_ID" and "TBLS"."TBL_NAME" = ? inner join "DBS" on "TBLS"."DB_ID" = "DBS"."DB_ID" and "DBS"."NAME" = ? inner join "PARTITION_KEY_VALS" "FILTER0" on "FILTER0"."PART_ID" = "PARTITIONS"."PART_ID" and "FILTER0"."INTEGER_IDX" = 0 inner join "PARTITION_KEY_VALS" "FILTER1" on "FILTER1"."PART_ID" = "PARTITIONS"."PART_ID" and "FILTER1"."INTEGER_IDX" = 1 where ( (((case when "TBLS"."TBL_NAME" = ? and "DBS"."NAME" = ? then cast("FILTER0"."PART_KEY_VAL" as decimal(21,0)) else null end) = ?) and ((case when "TBLS"."TBL_NAME" = ? and "DBS"."NAME" = ? then cast("FILTER1"."PART_KEY_VAL" as decimal(21,0)) else null end) = ?)) )".
at org.datanucleus.api.jdo.NucleusJDOHelper.getJDOExceptionForNucleusException(NucleusJDOHelper.java:451)
at org.datanucleus.api.jdo.JDOQuery.executeWithArray(JDOQuery.java:321)
at org.apache.hadoop.hive.metastore.MetaStoreDirectSql.getPartitionsViaSqlFilterInternal(MetaStoreDirectSql.java:300)
at org.apache.hadoop.hive.metastore.MetaStoreDirectSql.getPartitionsViaSqlFilter(MetaStoreDirectSql.java:211)
at org.apache.hadoop.hive.metastore.ObjectStore$3.getSqlResult(ObjectStore.java:1915)
at org.apache.hadoop.hive.metastore.ObjectStore$3.getSqlResult(ObjectStore.java:1909)
at org.apache.hadoop.hive.metastore.ObjectStore$GetHelper.run(ObjectStore.java:2208)
at org.apache.hadoop.hive.metastore.ObjectStore.getPartitionsByExprInternal(ObjectStore.java:1909)
at org.apache.hadoop.hive.metastore.ObjectStore.getPartitionsByExpr(ObjectStore.java:1882)

org.postgresql.util.PSQLException: ERROR: invalid input syntax for type numeric: "_DEFAULT_BINSRC_"

15/04/06 08:41:20 INFO metastore.ObjectStore: JDO filter pushdown cannot be used: Filtering is supported only on partition keys of type string
15/04/06 08:41:20 ERROR metastore.ObjectStore:
javax.jdo.JDOException: Exception thrown when executing query
at org.datanucleus.api.jdo.NucleusJDOHelper.getJDOExceptionForNucleusException(NucleusJDOHelper.java:596)
at org.datanucleus.api.jdo.JDOQuery.execute(JDOQuery.java:275)
at org.apache.hadoop.hive.metastore.ObjectStore.getPartitionNamesNoTxn(ObjectStore.java:1700)
at org.apache.hadoop.hive.metastore.ObjectStore.getPartitionNamesPrunedByExprNoTxn(ObjectStore.java:2003)
at org.apache.hadoop.hive.metastore.ObjectStore.access$400(ObjectStore.java:146)
at org.apache.hadoop.hive.metastore.ObjectStore$3.getJdoResult(ObjectStore.java:1937)
at org.apache.hadoop.hive.metastore.ObjectStore$3.getJdoResult(ObjectStore.java:1909)
at org.apache.hadoop.hive.metastore.ObjectStore$GetHelper.run(ObjectStore.java:2214)
at org.apache.hadoop.hive.metastore.ObjectStore.getPartitionsByExprInternal(ObjectStore.java:1909)
at org.apache.hadoop.hive.metastore.ObjectStore.getPartitionsByExpr(ObjectStore.java:1882)

Issue Links

• relates to: HIVE-8739 handle Derby and Oracle errors with joins and filters in Direct SQL in a invalid-DB-specific path
• links to: GitHub Pull Request #32
• links to: Review board