[
https://issues.apache.org/jira/browse/IMPALA-7028?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16487846#comment-16487846
]
Tim Armstrong commented on IMPALA-7028:
---------------------------------------
According to the profile, all of the Kudu scan nodes returned only a fraction
of the expected rows. I don't think the profile bug applies since it's a
successful DML, which waits for all the final status reports to be sent in.
{noformat}
18:26:14 E KUDU_SCAN_NODE (id=0):(Total: 4.999ms, non-child: 4.999ms,
% non-child: 100.00%)
18:26:14 E - BytesRead: 0
18:26:14 E - CollectionItemsRead: 0 (0)
18:26:14 E - KuduRemoteScanTokens: 0 (0)
18:26:14 E - NumScannerThreadsStarted: 1 (1)
18:26:14 E - PeakMemoryUsage: 12.00 KB (12288)
18:26:14 E - RowsRead: 10 (10)
18:26:14 E - RowsReturned: 10 (10)
18:26:14 E - RowsReturnedRate: 2.00 K/sec
18:26:14 E - ScanRangesComplete: 1 (1)
18:26:14 E - ScannerThreadsInvoluntaryContextSwitches: 4 (4)
18:26:14 E - ScannerThreadsTotalWallClockTime: 3.999ms
18:26:14 E - MaterializeTupleTime(*): 1.999ms
18:26:14 E - ScannerThreadsSysTime: 0.000ns
18:26:14 E - ScannerThreadsUserTime: 999.000us
18:26:14 E - ScannerThreadsVoluntaryContextSwitches: 3 (3)
18:26:14 E - TotalKuduScanRoundTrips: 1 (1)
18:26:14 E - TotalReadThroughput: 0.00 /sec
{noformat}
[~twmarshall] this looks like another case of Kudu scans mysteriously not
returning the expected number of rows. Are we tracking those with a single
JIRA. I guess the rows were inserted by the previous statement, run against the
same coordinator.
{noformat}
18:26:14 -- executing against localhost:21000
18:26:14 insert into tdata
18:26:14 select id, string_col, float_col, bigint_col, string_col, bool_col,
tinyint_col,
18:26:14 smallint_col, double_col, NULL, NULL, NULL from
functional_kudu.alltypes;
18:26:14
18:26:14 -- executing against localhost:21000
18:26:14 update tdata set vali = -1;
{noformat}
> TestKuduOperations.test_kudu_update fails due to wrong NumModifiedRows
> ----------------------------------------------------------------------
>
> Key: IMPALA-7028
> URL: https://issues.apache.org/jira/browse/IMPALA-7028
> Project: IMPALA
> Issue Type: Bug
> Components: Backend
> Affects Versions: Impala 3.1.0
> Reporter: Joe McDonnell
> Assignee: Tim Armstrong
> Priority: Critical
> Labels: broken-build, flaky
>
> Seen once on master exhaustive:
> {noformat}
> query_test/test_kudu.py:92: in test_kudu_update
> self.run_test_case('QueryTest/kudu_update', vector,
> use_db=unique_database)
> common/impala_test_suite.py:451: in run_test_case
> verify_runtime_profile(test_section['RUNTIME_PROFILE'],
> result.runtime_profile)
> common/test_result_verifier.py:590: in verify_runtime_profile
> actual))
> E AssertionError: Did not find matches for lines in runtime profile:
> E EXPECTED LINES:
> E NumModifiedRows: 7300
> ...
> # From the actual profile:
> E Partition: Default
> E NumModifiedRows: 32
> E NumRowErrors: 0{noformat}
> Will add a comment with the profile.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
