[IMPALA-2635] Hang on UNION query between 2 Kudu tables without computing stats

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: Kudu_Impala
• Fix Version/s: Kudu_Impala
• Component/s: Backend
• Labels: kudu
• Target Version: Impala 2.5.0
• Assignee: casey
• Reporter: Mike Percy
• Created: 04/Nov/15 23:14
• Updated: 29/Jan/16 05:50
• Resolved: 29/Jan/16 05:50

Description

This was all with Kudu_Impala and Kudu 0.5.0 beta.

I created 2 Kudu tables, one range partitioned and one hash partitioned, like so:

CREATE TABLE my_first_table (id BIGINT, name STRING) TBLPROPERTIES ('kudu.master_addresses'='host.example.com:7051', 'kudu.key_columns'='id', 'kudu.table_name'='my_first_table', 'storage_handler'='com.cloudera.kudu.hive.KuduStorageHandler');

CREATE TABLE test_table_1 (id BIGINT, name STRING)
DISTRIBUTE BY HASH(id) INTO 16 BUCKETS
TBLPROPERTIES ( 'kudu.master_addresses'='host1.example.com:7051', 'kudu.key_columns'='id', 'kudu.table_name'='test_table_1', 'storage_handler'='com.cloudera.kudu.hive.KuduStorageHandler');

Then insert 2 rows:

INSERT INTO my_first_table (id, name) VALUES (0, 'Foo');
INSERT INTO test_table_1 (id, name) VALUES (1, 'Blah');

Now when I do a normal COUNT or SELECT * everything is fine:

> SELECT * FROM my_first_table;
Query: select * FROM my_first_table
+----+------+
| id | name |
+----+------+
| 0  | Foo  |
+----+------+
Fetched 1 row(s) in 0.30s

> select * from test_table_1;
Query: select * from test_table_1
+----+------+
| id | name |
+----+------+
| 1  | Blah |
+----+------+

However when I attempt to do a UNION it hangs forever:

> SELECT * from my_first_table UNION SELECT * from test_table_1;
Query: select * from my_first_table UNION SELECT * from test_table_1

Doing a COMPUTE STATS on each table prevents the hang, but that should not be required.