[
https://issues.apache.org/jira/browse/IMPALA-5999?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Bikramjeet Vig resolved IMPALA-5999.
------------------------------------
Resolution: Fixed
Fix Version/s: Impala 2.11.0
https://github.com/apache/incubator-impala/commit/1ca3adf46c5ef5055c13fd3ce57e7c53218c219c
Some test runs are currently failing randomly in test_ir_functions due
to LLVM linkage error. This happens when impala tries to link a function
from a lib file on hdfs, but it somehow got removed from the lib cache
before it could link. This results in a new file being cached but with
a slightly different local filename, and since impala only uses local
filenames to check for collision for linking of LLVM modules, it ends
up linking the same file twice and hence encounters an error.
This patch fixes this issue by using the hdfs file path to check for
collision of lib files.
Testing:
Added a backend test that tries to link the same lib twice.
Change-Id: Iefa23bbf47998fe7e84011e1edf8e794e94a1757
Reviewed-on: http://gerrit.cloudera.org:8080/8487
Reviewed-by: Tim Armstrong <[email protected]>
Tested-by: Impala Public Jenkins
> Multiple failures in TestUdfExecution and test_spilling
> -------------------------------------------------------
>
> Key: IMPALA-5999
> URL: https://issues.apache.org/jira/browse/IMPALA-5999
> Project: IMPALA
> Issue Type: Bug
> Components: Backend
> Affects Versions: Impala 2.11.0
> Reporter: Sailesh Mukil
> Assignee: Bikramjeet Vig
> Priority: Blocker
> Labels: broken-build, flaky-test
> Fix For: Impala 2.11.0
>
>
> We've seen 3 instances of failed GVOs with the same errors:
> https://jenkins.impala.io/job/ubuntu-16.04-from-scratch/353/consoleFull
> https://jenkins.impala.io/job/ubuntu-16.04-from-scratch/346/consoleFull
> https://jenkins.impala.io/job/ubuntu-16.04-from-scratch/361/consoleFull
> {code:java}
> 03:40:20 [gw14] FAILED
> query_test/test_udfs.py::TestUdfExecution::test_java_udfs[exec_option:
> {'disable_codegen_rows_threshold': 0, 'disable_codegen': True,
> 'exec_single_node_rows_threshold': 100, 'enable_expr_rewrites': True} |
> table_format: text/none]
> 03:40:22 [gw8] FAILED
> query_test/test_udfs.py::TestUdfExecution::test_java_udfs[exec_option:
> {'disable_codegen_rows_threshold': 0, 'disable_codegen': True,
> 'exec_single_node_rows_threshold': 0, 'enable_expr_rewrites': False} |
> table_format: text/none]
> 03:40:26 [gw1] FAILED
> query_test/test_udfs.py::TestUdfExecution::test_native_functions[exec_option:
> {'disable_codegen_rows_threshold': 0, 'disable_codegen': True,
> 'exec_single_node_rows_threshold': 0, 'enable_expr_rewrites': True} |
> table_format: text/none]
> 03:40:26 [gw9] FAILED
> query_test/test_nested_types.py::TestNestedTypes::test_tpch[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: parquet/none]
> 03:40:26
> query_test/test_nested_types.py::TestNestedTypes::test_parquet_stats[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: parquet/none]
> 03:40:26 [gw7] FAILED
> query_test/test_scanners.py::TestScannersAllTableFormatsWithLimit::test_limit[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: parquet/none]
> 03:40:26
> query_test/test_scanners.py::TestScannersAllTableFormatsWithLimit::test_limit[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: avro/snap/block]
> 03:40:26 [gw0] FAILED
> query_test/test_spilling.py::TestSpillingDebugActionDimensions::test_spilling[exec_option:
> {'debug_action': '-1:OPEN:[email protected]',
> 'default_spillable_buffer_size': '256k'} | table_format: parquet/none]
> ....
> {code}
> (More useful logs in the links above, too large to paste)
> It seems to be flaky as it doesn't seem to show up every time.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
