You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues-all@impala.apache.org by "Tim Armstrong (JIRA)" <ji...@apache.org> on 2018/08/03 20:46:00 UTC

[jira] [Comment Edited] (IMPALA-7378) test_strict_mode failed on an ASAN build: expected "Error converting column: 5 to DOUBLE"

    [ https://issues.apache.org/jira/browse/IMPALA-7378?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16568732#comment-16568732 ] 

Tim Armstrong edited comment on IMPALA-7378 at 8/3/18 8:45 PM:
---------------------------------------------------------------

That instance was similar. It seems that there's a bug in status propagation from HdfsScanNode::ProcessSplit() (where we logged  "Error parsing row:") to FragmentInstanceState::Exec() which returned the OK status logged by QueryStatus::ExecFInstance() -  "status=OK".
{noformat}
I0803 02:03:23.043638 26371 status.cc:125] Error converting column: 0 to TINYINT
    @          0x21c9d17
    @          0x17560c4
    @          0x24223bb
    @          0x2421dbb
    @     0x7f012f368b07
    @          0x247dff1
    @          0x247f81c
    @          0x24810d9
    @          0x241a49d
    @          0x23d87a9
    @          0x23d78b7
    @          0x1c40f07
    @          0x22d1cef
    @          0x22dcee9
    @          0x22dcd3c
    @          0x384131a
    @     0x7f0592839e25
    @     0x7f059235034d
I0803 02:03:23.043845 26371 runtime-state.cc:196] Error from query 56419b1e2575fc0a:4f79b50100000000: Error parsing row: file: hdfs://localhost:20500/test-warehouse/overflow/overflow.txt, before offset: 454
I0803 02:03:23.043987 26371 hdfs-scan-node.cc:498] Scan node (id=0) ran into a parse error for scan range hdfs://localhost:20500/test-warehouse/overflow/overflow.txt(0:454). Processed 454 bytes.
I0803 02:03:23.085072 26366 query-state.cc:405] Instance completed. instance_id=56419b1e2575fc0a:4f79b50100000001 #in-flight=3 status=OK
{noformat}
I think we should consider adding some additional logging (maybe only enabled under ASAN) that will let us bisect where the status is dropped.


was (Author: tarmstrong):
That instance was similar. It seems that there's a bug in status propagation from HdfsScanNode::ProcessSplit() (where we logged  "Error parsing row:") to FragmentInstanceState::Exec() which returned the OK status logged by QueryStatus::ExecFInstance() -  "status=OK".

I think we should consider adding some additional logging (maybe only enabled under ASAN) that will let us bisect where the status is dropped.

> test_strict_mode failed on an ASAN build: expected "Error converting column: 5 to DOUBLE"
> -----------------------------------------------------------------------------------------
>
>                 Key: IMPALA-7378
>                 URL: https://issues.apache.org/jira/browse/IMPALA-7378
>             Project: IMPALA
>          Issue Type: Bug
>          Components: Backend
>    Affects Versions: Impala 3.1.0
>            Reporter: David Knupp
>            Assignee: Tim Armstrong
>            Priority: Blocker
>              Labels: broken-build
>
> *Stacktrace*
> {noformat}
> query_test/test_queries.py:159: in test_strict_mode
>     self.run_test_case('QueryTest/strict-mode-abort', vector)
> common/impala_test_suite.py:420: in run_test_case
>     assert False, "Expected exception: %s" % expected_str
> E   AssertionError: Expected exception: Error converting column: 5 to DOUBLE
> {noformat}
> *Standard Error*
> {noformat}
> -- executing against localhost:21000
> use functional;
> SET strict_mode=1;
> SET batch_size=0;
> SET num_nodes=0;
> SET disable_codegen_rows_threshold=0;
> SET disable_codegen=False;
> SET abort_on_error=0;
> SET exec_single_node_rows_threshold=0;
> -- executing against localhost:21000
> select * from overflow;
> -- executing against localhost:21000
> use functional;
> SET strict_mode=1;
> SET batch_size=0;
> SET num_nodes=0;
> SET disable_codegen_rows_threshold=0;
> SET disable_codegen=False;
> SET abort_on_error=1;
> SET exec_single_node_rows_threshold=0;
> -- executing against localhost:21000
> select tinyint_col from overflow;
> -- executing against localhost:21000
> select smallint_col from overflow;
> -- executing against localhost:21000
> select int_col from overflow;
> -- executing against localhost:21000
> select bigint_col from overflow;
> -- executing against localhost:21000
> select float_col from overflow;
> -- executing against localhost:21000
> select double_col from overflow;
> {noformat}



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-all-unsubscribe@impala.apache.org
For additional commands, e-mail: issues-all-help@impala.apache.org