[IMPALA-11307] Loading snapshot testdata failed by cannot dropping tables used by MV

• Type: Bug
• Status: Open
• Priority: Critical
• Resolution: Unresolved
• Component/s: Infrastructure
• Epic Link: Materialized View
• Epic Color: ghx-label-9
• Assignee: Quanlong Huang
• Reporter: Quanlong Huang
• Created: 20/May/22 07:16
• Updated: 24/Jun/22 09:34

Description

I tried to load a snapshot of testdata and encountered this failure:

Loading TPC-H data (logging to /data/jenkins/workspace/impala-private-parameterized/repos/Impala/logs/data_loading/load-tpch.log)... 
Loading TPC-DS data (logging to /data/jenkins/workspace/impala-private-parameterized/repos/Impala/logs/data_loading/load-tpcds.log)... 
    FAILED (Took: 4 min 52 sec)
    'load-data functional-query exhaustive' failed. Tail of log:
21:30:39 load-functional-query-exhaustive-hive-generated-seq-bzip-record.sql
21:30:39 load-functional-query-exhaustive-hive-generated-seq-snap-block.sql
21:30:39 load-functional-query-exhaustive-hive-generated-seq-snap-record.sql
...
21:32:22 Error executing impala SQL: /data/jenkins/workspace/impala-private-parameterized/repos/Impala/logs/data_loading/sql/functional/create-functional-query-exhaustive-impala-generated-seq-snap-record.sql See: /data/jenkins/workspace/impala-private-parameterized/repos/Impala/logs/data_loading/sql/functional/create-functional-query-exhaustive-impala-generated-seq-snap-record.sql.log
21:32:22 Beginning execution of impala SQL on localhost: /data/jenkins/workspace/impala-private-parameterized/repos/Impala/logs/data_loading/sql/functional/create-functional-query-exhaustive-impala-generated-rc-gzip-block.sql
Background task Loading functional-query data (pid 8863) failed.
  Loading workload 'tpch' using exploration strategy 'core' OK (Took: 6 min 25 sec)
  Loading workload 'tpcds' using exploration strategy 'core' OK (Took: 11 min 42 sec)
ERROR in /data/jenkins/workspace/impala-private-parameterized/repos/Impala/testdata/bin/create-load-data.sh at line 85: fi
Generated: /data/jenkins/workspace/impala-private-parameterized/repos/Impala/logs/extra_junit_xml_logs/generate_junitxml.buildall.create-load-data.20220520_04_39_12.xml

Looking into the error log file, the failure is

ERROR: DROP TABLE IF EXISTS functional_seq_record_snap.insert_only_transactional_table
Traceback (most recent call last):
  File "/data/jenkins/workspace/impala-private-parameterized/repos/Impala/bin/load-data.py", line 188, in exec_impala_query_from_file
    result = impala_client.execute(query)
  File "/data/jenkins/workspace/impala-private-parameterized/repos/Impala/tests/beeswax/impala_beeswax.py", line 189, in execute
    handle = self.__execute_query(query_string.strip(), user=user)
  File "/data/jenkins/workspace/impala-private-parameterized/repos/Impala/tests/beeswax/impala_beeswax.py", line 367, in __execute_query
    self.wait_for_finished(handle)
  File "/data/jenkins/workspace/impala-private-parameterized/repos/Impala/tests/beeswax/impala_beeswax.py", line 388, in wait_for_finished
    raise ImpalaBeeswaxException("Query aborted:" + error_log, None)
ImpalaBeeswaxException: ImpalaBeeswaxException:
 Query aborted:ImpalaRuntimeException: Error making 'dropTable' RPC to Hive Metastore: 
CAUSED BY: MetaException: Cannot drop table as it is used in the following materialized views [functional_seq_record_snap.materialized_view]

The DROP TABLE statement is executed before each CREATE TABLE statment, e.g.

DROP TABLE IF EXISTS functional_seq_record_snap.insert_only_transactional_table;
CREATE  TABLE IF NOT EXISTS functional_seq_record_snap.insert_only_transactional_table (
col1 int
)
STORED AS SEQUENCEFILE
LOCATION '/test-warehouse/managed/functional_seq_record_snap.db/insert_only_transactional_table_seq_record_snap'
TBLPROPERTIES (
'transactional_properties' = 'insert_only',
'transactional' = 'true'
);

The MV is supposed to be created later in load-functional-query-exhaustive-hive-generated-seq-snap-record.sql:

DROP TABLE IF EXISTS functional_seq_record_snap.materialized_view;
-- The create materialized view command is moved down so that the database's
-- managed directory has been created. Otherwise the command would fail. This
-- is a bug in Hive.
CREATE MATERIALIZED VIEW IF NOT EXISTS functional_seq_record_snap.materialized_view
  AS SELECT * FROM functional_seq_record_snap.insert_only_transactional_table;

As I'm loading the snapshot which is done by loading HMS snapshot and then loading the HDFS snapshot, I think those CREATE/DROP statements should be skipped.

CC amansinha