[IMPALA-6331] slow s3 operations in unit tests

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Cannot Reproduce
• Affects Version/s: Impala 2.10.0
• Fix Version/s: Not Applicable
• Component/s: Backend, Infrastructure, Perf Investigation
• Epic Color: ghx-label-1
• Assignee: Unassigned
• Reporter: Vuk Ercegovac
• Created: 15/Dec/17 20:30
• Updated: 04/Dec/20 19:11
• Resolved: 04/Dec/20 19:11

Description

Seeing some slowness for certain fs tests when run against S3. Since they in the hour range, I think its useful to track it down and distinguish between a test infrastructure issue vs. an impala/s3 issue.
Here are logs from a recent run (note the time gaps in the log entries):

20:07:37 query_test/test_hdfs_file_mods.py::TestHdfsFileMods::test_file_modifications[modification_type: move_file | exec_option:

{'batch_size': 0, 'num_nodes': 0, 'disable_codegen_rows_threshold': 0, 'disable_codegen': False, 'abort_on_error': 1, 'exec_single_node_rows_threshold': 0}

| table_format: text/none] PASSED
20:08:19 query_test/test_hdfs_file_mods.py::TestHdfsFileMods::test_file_modifications[modification_type: append | exec_option:

{'batch_size': 0, 'num_nodes': 0, 'disable_codegen_rows_threshold': 0, 'disable_codegen': False, 'abort_on_error': 1, 'exec_single_node_rows_threshold': 0}

| table_format: text/none] PASSED
22:59:55 query_test/test_insert.py::TestInsertQueries::test_insert[exec_option:

{'sync_ddl': 0, 'batch_size': 0, 'num_nodes': 0, 'disable_codegen_rows_threshold': 0, 'disable_codegen': True, 'abort_on_error': 1, 'exec_single_node_rows_threshold': 0}

| table_format: text/none] PASSED
03:40:11 query_test/test_insert.py::TestInsertQueries::test_insert[exec_option:

{'sync_ddl': 0, 'batch_size': 0, 'num_nodes': 0, 'disable_codegen_rows_threshold': 0, 'disable_codegen': False, 'abort_on_error': 1, 'exec_single_node_rows_threshold': 0}

| table_format: text/none] PASSED
03:41:12 query_test/test_insert.py::TestInsertQueries::test_insert_overwrite[exec_option:

{'sync_ddl': 0, 'batch_size': 0, 'num_nodes': 0, 'disable_codegen_rows_threshold': 0, 'disable_codegen': True, 'abort_on_error': 1, 'exec_single_node_rows_threshold': 0}

| table_format: text/none] PASSED

There are several other tests that take in the 10 minute to minute range. Highlighting above the longest ones (> hour).