[
https://issues.apache.org/jira/browse/IMPALA-11803?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17694784#comment-17694784
]
ASF subversion and git services commented on IMPALA-11803:
----------------------------------------------------------
Commit fd3bccf5f4b30bc3bd754dc11f41f6629047f5b1 in impala's branch
refs/heads/master from Joe McDonnell
[ https://gitbox.apache.org/repos/asf?p=impala.git;h=fd3bccf5f ]
IMPALA-11803: Fix hitting DCHECK when running union on empty table with MT_DOP>1
The DCHECK was caused as the value of useMtScanNode_
was not getting set for empty tables. This was because of
a condition for empty tables that returned from the
computeNodeResourceProfile() function before setting up
useMtScanNode_. It got rectified by placing the
assignment of useMtScanNode_ before that condition.
Testing: The fix is verified with an end-to-end test in test_mt_dop.py.
Change-Id: Idbae5e1a78211327a214b2d936743bda767ae3c4
Reviewed-on: http://gerrit.cloudera.org:8080/19474
Reviewed-by: Impala Public Jenkins <[email protected]>
Tested-by: Impala Public Jenkins <[email protected]>
> impalad hit DCHECK when running union on empty table with MT_DOP>1
> ------------------------------------------------------------------
>
> Key: IMPALA-11803
> URL: https://issues.apache.org/jira/browse/IMPALA-11803
> Project: IMPALA
> Issue Type: Bug
> Reporter: Quanlong Huang
> Assignee: Pranav Yogi Lodha
> Priority: Major
>
> This is reported from a community user. When using the DEBUG build, impalad
> could hit DCHECK and crash when running union with empty table with MT_DOP>1.
> I can reproduce the issue by launching the impala cluster with only one node:
> {code:java}
> bin/start-impala-cluster.py -r -s 1
> {code}
> Then run the following queries
> {code:sql}
> create table empty_tbl (id int) stored as parquet;
> create table tbl (id int) stored as parquet;
> insert into tbl values (0), (1), (2);
> insert into tbl values (3);
> set mt_dop=2;
> select count(*) from (select * from empty_tbl union all select * from tbl) t;
> {code}
> The stacktrace in impalad.ERROR:
> {code:java}
> F1216 14:51:44.285214 24142 hdfs-scan-node-base.cc:352]
> b444e7823f021f02:5af2aa4700000000] Check failed:
> shared_state_.use_mt_scan_node_ || instance_ctx_pbs.size() == 1 Non MT scan
> node should only have a single instance.
> *** Check failure stack trace: ***
> @ 0x379c53d google::LogMessage::Fail()
> @ 0x379e474 google::LogMessage::SendToLog()
> @ 0x379bf1c google::LogMessage::Flush()
> @ 0x379e999 google::LogMessageFatal::~LogMessageFatal()
> @ 0x197ddfd
> impala::HdfsScanPlanNode::ProcessScanRangesAndInitSharedState()
> @ 0x197ef4f impala::HdfsScanPlanNode::Init()
> @ 0x1961150
> _ZN6impala8PlanNode16CreateTreeHelperEPNS_13FragmentStateERKSt6vectorINS_9TPlanNodeESaIS4_EEPS0_PiPS9_.localalias
> @ 0x1961082
> _ZN6impala8PlanNode16CreateTreeHelperEPNS_13FragmentStateERKSt6vectorINS_9TPlanNodeESaIS4_EEPS0_PiPS9_.localalias
> @ 0x1961082
> _ZN6impala8PlanNode16CreateTreeHelperEPNS_13FragmentStateERKSt6vectorINS_9TPlanNodeESaIS4_EEPS0_PiPS9_.localalias
> @ 0x1961205 impala::PlanNode::CreateTree()
> @ 0x13c56c2 impala::FragmentState::Init()
> @ 0x13c76b9 impala::FragmentState::CreateFragmentStateMap()
> @ 0x134492e impala::QueryState::StartFInstances()
> @ 0x1335036 impala::QueryExecMgr::ExecuteQueryHelper()
> @ 0x1336100
> boost::detail::function::void_function_obj_invoker0<>::invoke()
> @ 0x18c7018 impala::Thread::SuperviseThread()
> @ 0x18c7e21 boost::detail::thread_data<>::run()
> @ 0x239bb77 thread_proxy
> @ 0x7effc4fcf6db start_thread
> @ 0x7effc1d6f61f clone
> {code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
