[Bug] vectorized compaction failed

[hello-stephen]

Search before asking

• I had searched in the issues and found no similar issues.

Version

branch: master
commit id: 953429e

What's Wrong?

4 server, 1fe 3be deploy,
I use tools/tpch-tools to load 100G data, after it finished load 1 hour later, I check it and find that lineitem table failed to do cumulative compaction. Then I found error log in be.WARNING ,

W0516 13:37:26.612212 801433 compaction.cpp:252] row_num does not match between cumulative input and output! input_row_num=18742298, merged_row_num=0, filtered_row_num=0, output_row_num=0
W0516 13:37:26.612402 801433 tablet.cpp:1410] failed to do cumulative compaction. res=Internal error: :  0# doris::Status::ConstructErrorStatus(short, doris::Slice const&) at /mnt/hdd01/ldy/incubator-doris/be/src/common/status.cpp:80
 1# doris::Compaction::check_correctness(doris::Merger::Statistics const&) at /mnt/hdd01/ldy/incubator-doris/be/src/olap/compaction.cpp:276
 2# doris::Compaction::do_compaction_impl(long) at /mnt/hdd01/ldy/incubator-doris/be/src/olap/compaction.cpp:121
 3# doris::Compaction::do_compaction(long) at /mnt/hdd01/ldy/incubator-doris/be/src/olap/compaction.cpp:57
 4# doris::CumulativeCompaction::execute_compact_impl() at /mnt/hdd01/ldy/incubator-doris/be/src/olap/cumulative_compaction.cpp:75
 5# doris::Compaction::execute_compact() at /mnt/hdd01/ldy/incubator-doris/be/src/olap/compaction.cpp:46
 6# doris::Tablet::execute_compaction(doris::CompactionType) at /mnt/hdd01/ldy/incubator-doris/be/src/olap/tablet.cpp:1407
 7# std::_Function_handler<void (), doris::StorageEngine::_submit_compaction_task(std::shared_ptr<doris::Tablet>, doris::CompactionType)::{lambda()#1}>::_M_invoke(std::_Any_data const&) at /var/local/ldb-toolchain/include/c++/11/bits/std_function.h:291
 8# doris::ThreadPool::dispatch_thread() at /mnt/hdd01/ldy/incubator-doris/be/src/util/threadpool.cpp:548
 9# doris::Thread::supervise_thread(void*) at /mnt/hdd01/ldy/incubator-doris/be/src/util/thread.cpp:409
10# start_thread in /lib/x86_64-linux-gnu/libpthread.so.0
11# __clone in /lib/x86_64-linux-gnu/libc.so.6
 (error -224), tablet=1007524.1617673526.414fcdf78976e781-857a7348cc73c68e

This error log still keep printing after I drop lineitem and drop database.
then I set enable_vectorized_compaction=false in be.conf and restart cluster, it stops print.
At this time, it looks like recovered, so I redo load again, and no error log again.

To summarize：
Load tpch 100G data with enable_vectorized_compaction=true, NOT OK;
Load tpch 100G data with enable_vectorized_compaction=false, OK;

What You Expected?

do cumulative compaction and base compaction successful as expected.

How to Reproduce?

use tools/tpch-tools to load 100G data.

Anything Else?

If more information is required, I will provide it

Are you willing to submit PR?

• Yes I am willing to submit a PR!

Code of Conduct

• I agree to follow this project's Code of Conduct