[ 
https://issues.apache.org/jira/browse/ARROW-15793?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17498886#comment-17498886
 ] 

David Li commented on ARROW-15793:
----------------------------------

I investigated this here and it seems to be a gRPC issue: 
https://github.com/apache/arrow/pull/12499/files#diff-2d05b3bd0ce469f82d903cb7f040168dea4e1b4713e4fec54fed2fa7842ec6e9R620

{noformat}
  // For gRPC: This test randomly takes 10 seconds because gRPC has problems
  // joining its internal threads. It's because it's waiting on a thread
  // calling epoll...with a timeout of 10 seconds.
  // https://groups.google.com/g/grpc-io/c/j7xmW0F3eu0
  // "Grpc Client-side C++ Channel Destructor taking 10 seconds"
{noformat}

Though maybe we can remove the {{impl_.reset()}} call from FlightClient, I'll 
give that a try.

> [C++][FlightRPC] DoPutLargeBatch test sometimes stuck for 10 seconds
> --------------------------------------------------------------------
>
>                 Key: ARROW-15793
>                 URL: https://issues.apache.org/jira/browse/ARROW-15793
>             Project: Apache Arrow
>          Issue Type: Bug
>          Components: C++, FlightRPC
>            Reporter: Yibo Cai
>            Priority: Major
>
> Normally the test finishes in 100ms. But it often costs 10s on my test 
> machine.
> Debug build is good.
> I did brief debug, looks it's related to 
> [https://github.com/apache/arrow/pull/12302].
> It stuck 10 seconds in destructing grpc::Server at code 
> [https://github.com/apache/arrow/blob/master/cpp/src/arrow/flight/server.cc#L863]
> To reproduce:
> {code:bash}
> $ cmake -GNinja -DARROW_BUILD_TESTS=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo 
> -DARROW_FLIGHT=ON ..
> $ ninja arrow-flight-test
> $ relwithdebinfo/arrow-flight-test --gtest_filter="*DoPutLargeBatch*"
> [==========] Running 1 test from 1 test suite.
> [----------] Global test environment set-up.
> [----------] 1 test from TestDoPut
> [ RUN      ] TestDoPut.DoPutLargeBatch
> [       OK ] TestDoPut.DoPutLargeBatch (10017 ms)
> [----------] 1 test from TestDoPut (10017 ms total)
> [----------] Global test environment tear-down
> [==========] 1 test from 1 test suite ran. (10017 ms total)
> [  PASSED  ] 1 test.
> {code}



--
This message was sent by Atlassian Jira
(v8.20.1#820001)

Reply via email to