This is an automated email from the ASF dual-hosted git repository.
dataroaring pushed a commit to branch branch-3.0
in repository https://gitbox.apache.org/repos/asf/doris.git
The following commit(s) were added to refs/heads/branch-3.0 by this push:
new 4c3acd7ed26 [fix](move-memtable) do not execute close if create rowset
failed when loading MOW table (#40105)
4c3acd7ed26 is described below
commit 4c3acd7ed2671c2036b5467a8e58734c50f0e6bb
Author: hui lai <1353307...@qq.com>
AuthorDate: Mon Sep 2 09:50:18 2024 +0800
[fix](move-memtable) do not execute close if create rowset failed when
loading MOW table (#40105)
Core dump happened when load to MOW table:
```
Check failure stack trace: ***
@ 0x55fae437d246 google::LogMessage::SendToLog()
@ 0x55fae4379c90 google::LogMessage::Flush()
@ 0x55fae437da89 google::LogMessageFatal::~LogMessageFatal()
@ 0x55faacf26bbf doris::BaseTablet::check_delete_bitmap_correctness()
@ 0x55fab05049ef doris::RowsetBuilder::commit_txn()
@ 0x55fab09026e8 doris::LoadStreamWriter::close()
@ 0x55fab089eff7 std::_Function_handler<>::_M_invoke()
@ 0x55fab0d14d7c doris::WorkThreadPool<>::work_thread()
@ 0x55fae76ae6f0 execute_native_thread_routine
@ 0x7fa32ea45ac3 (unknown)
@ 0x7fa32ead7850 (unknown)
@ (nil) (unknown)
Query id: a21981d5c8ef4113-84df9a5a8680e004 ***
is nereids: 0 ***
tablet id: 0 ***
Aborted at 1724668499 (unix time) try "date -d @1724668499" if you are
using GNU date ***
Current BE git commitID: 2f848737c1 ***
SIGABRT unknown detail explain (@0x20db) received by PID 8411 (TID 9837 OR
0x7f9e42cfe640) from PID 8411; stack trace: ***
0# doris::signal::(anonymous namespace)::FailureSignalHandler(int,
siginfo_t*, void*) at
/home/zcp/repo_center/doris_master/doris/be/src/common/signal_handler.h:421
1# 0x00007FA32E9F3520 in /lib/x86_64-linux-gnu/libc.so.6
2# pthread_kill at ./nptl/pthread_kill.c:89
3# raise at ../sysdeps/posix/raise.c:27
4# abort at ./stdlib/abort.c:81
5# 0x000055FAE4387B1D in /mnt/hdd01/ci/master-deploy/be/lib/doris_be
6# 0x000055FAE437A15A in /mnt/hdd01/ci/master-deploy/be/lib/doris_be
7# google::LogMessage::SendToLog() in
/mnt/hdd01/ci/master-deploy/be/lib/doris_be
8# google::LogMessage::Flush() in
/mnt/hdd01/ci/master-deploy/be/lib/doris_be
9# google::LogMessageFatal::~LogMessageFatal() in
/mnt/hdd01/ci/master-deploy/be/lib/doris_be
10# doris::BaseTablet::check_delete_bitmap_correctness(std::shared_ptr,
long, long, std::unordered_set, std::equal_to, std::allocator > const&,
std::vector, std::allocator > >*) at
/home/zcp/repo_center/doris_master/doris/be/src/olap/base_tablet.cpp:1152
11# doris::RowsetBuilder::commit_txn() at
/home/zcp/repo_center/doris_master/doris/be/src/olap/rowset_builder.cpp:316
12# doris::LoadStreamWriter::close() at
/home/zcp/repo_center/doris_master/doris/be/src/runtime/load_stream_writer.cpp:311
13# std::_Function_handler::_M_invoke(std::_Any_data const&) at
/var/local/ldb-toolchain/bin/../lib/gcc/x86_64-linux-gnu/11/../../../../include/c++/11/bits/std_function.h:291
14# doris::WorkThreadPool::work_thread(int) at
/home/zcp/repo_center/doris_master/doris/be/src/util/work_thread_pool.hpp:159
15# execute_native_thread_routine at
../../../../../libstdc+-v3/src/c+11/thread.cc:84
16# start_thread at ./nptl/pthread_create.c:442
17# 0x00007FA32EAD7850 at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:83
```
If create rowset failed,` calc_delete_bitmap_task` still could be
executed:
```
add segment failed load_id=5649413b98976f0d-a105b42749f561b0, txn_id=2,
tablet_id=10088, status=[INTERNAL_ERROR]create row
set failed
...
submit calc delete bitmap task to executor, tablet_id: 10088, txn_id: 2
```
This PR skips close to avoid `submit_calc_delete_bitmap_task` if create
rowset failed when loading MOW table to solve this problem.
---
be/src/runtime/load_stream.cpp | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/be/src/runtime/load_stream.cpp b/be/src/runtime/load_stream.cpp
index aa1749caace..066d81cc17d 100644
--- a/be/src/runtime/load_stream.cpp
+++ b/be/src/runtime/load_stream.cpp
@@ -281,6 +281,13 @@ Status TabletStream::close() {
return _status;
}
+ // it is necessary to check status after wait_func,
+ // for create_rowset could fail during add_segment when loading to MOW
table,
+ // in this case, should skip close to avoid submit_calc_delete_bitmap_task
which could cause coredump.
+ if (!_status.ok()) {
+ return _status;
+ }
+
auto close_func = [this, &mu, &cv]() {
signal::set_signal_task_id(_load_id);
auto st = _load_stream_writer->close();
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@doris.apache.org
For additional commands, e-mail: commits-h...@doris.apache.org
