[Fix](hdfs-writer) Fix hdfs file writer core with check failed: _ref_cnt == 0 in dtor of HdfsFileWriter.
#33959
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…_cnt == 0` in dctor of `HdfsFileWriter`.
|
Thank you for your contribution to Apache Doris. Since 2024-03-18, the Document has been moved to doris-website. |
Contributor
Author
|
run buildall |
Contributor
|
clang-tidy review says "All clean, LGTM! 👍" |
|
TeamCity be ut coverage result: |
TPC-H: Total hot run time: 38297 ms |
TPC-DS: Total hot run time: 185925 ms |
ClickBench: Total hot run time: 30.02 s |
gavinchou
reviewed
Apr 22, 2024
| if (!handle->invalid()) { | ||
| handle->inc_ref(); | ||
| *fs_handle = handle; | ||
| *fs_handle = std::move(handle); |
Contributor
There was a problem hiding this comment.
why do we still have to deal with inc_ref() even with shared_ptr support?
can we integrate the process (inc_ref, dec_ref) into shared_ptr ?
Contributor
Author
There was a problem hiding this comment.
It is for fs handle cache.
platoneko
requested changes
Apr 23, 2024
be/src/io/fs/hdfs_file_writer.cpp
Outdated
| HdfsFileWriter::HdfsFileWriter(Path path, std::shared_ptr<HdfsHandler> handler, hdfsFile hdfs_file, | ||
| std::string fs_name, const FileWriterOptions* opts) | ||
| : _path(std::move(path)), | ||
| _hdfs_handler(handler), |
Contributor
There was a problem hiding this comment.
Suggested change
| _hdfs_handler(handler), | |
| _hdfs_handler(std::move(handler)), |
Contributor
There was a problem hiding this comment.
ditto:
return std::make_unique<HdfsFileWriter>(std::move(path), std::move(handler), hdfs_file, fs_name, opts);
| // do not use std::shared_ptr or std::unique_ptr | ||
| // _fs_handle is managed by HdfsFileSystemCache | ||
| HdfsHandler* _fs_handle = nullptr; | ||
| std::shared_ptr<HdfsHandler> _fs_handle = nullptr; |
Contributor
There was a problem hiding this comment.
Suggested change
| std::shared_ptr<HdfsHandler> _fs_handle = nullptr; | |
| std::shared_ptr<HdfsHandler> _fs_handle; |
可以不用显式写 = nullptr，智能指针默认初始化都是 nullptr
be/src/io/hdfs_util.cpp
Outdated
| } | ||
| if (_cache.size() < MAX_CACHE_HANDLE) { | ||
| std::unique_ptr<HdfsHandler> handle = std::make_unique<HdfsHandler>(hdfs_fs, true); | ||
| std::shared_ptr<HdfsHandler> handle = std::make_shared<HdfsHandler>(hdfs_fs, true); |
Contributor
There was a problem hiding this comment.
Suggested change
| std::shared_ptr<HdfsHandler> handle = std::make_shared<HdfsHandler>(hdfs_fs, true); | |
| auto handle = std::make_shared<HdfsHandler>(hdfs_fs, true); |
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
| _last_access_time = std::chrono::duration_cast<std::chrono::milliseconds>( | ||
| std::chrono::system_clock::now().time_since_epoch()) | ||
| .count(); | ||
| void update_last_access_time() { |
Contributor
There was a problem hiding this comment.
warning: method 'update_last_access_time' can be made const [readability-make-member-function-const]
Suggested change
| void update_last_access_time() { | |
| void update_last_access_time() const { |
kaka11chen
force-pushed
the
fix_hdfs_file_writer_core
branch
2 times, most recently
from
60e60ff to
ffc4826
Compare
kaka11chen
force-pushed
the
fix_hdfs_file_writer_core
branch
from
ffc4826 to
c95dd69
Compare
Contributor
Author
|
run buildall |
|
TeamCity be ut coverage result: |
platoneko
approved these changes
Apr 23, 2024
github-actions
bot
added
the
approved
Contributor
|
PR approved by at least one committer and no changes requested. |
Contributor
|
PR approved by anyone and no changes requested. |
dataroaring
pushed a commit
that referenced
this pull request
Apr 23, 2024
…_cnt == 0` in dtor of `HdfsFileWriter`. (#33959) ## Issue: ``` F20240421 17:14:37.494115 184986 hdfs_util.h:65] Check failed: _ref_cnt == 0 *** Check failure stack trace: *** F20240421 17:14:37.505879 185108 hdfs_util.h:65] Check failed: _ref_cnt == 0 *** Check failure stack trace: *** @ 0x556f5236d316 google::LogMessageFatal::~LogMessageFatal() @ 0x556f5236d316 google::LogMessageFatal::~LogMessageFatal() @ 0x556f2830e200 doris::io::HdfsFileWriter::~HdfsFileWriter() @ 0x556f2830e200 doris::io::HdfsFileWriter::~HdfsFileWriter() @ 0x556f2830e21e doris::io::HdfsFileWriter::~HdfsFileWriter() @ 0x556f2830e21e doris::io::HdfsFileWriter::~HdfsFileWriter() @ 0x556f507893b0 doris::vectorized::VHivePartitionWriter::~VHivePartitionWriter() @ 0x556f507893b0 doris::vectorized::VHivePartitionWriter::~VHivePartitionWriter() @ 0x556f506c005e std::_Hashtable<>::clear() @ 0x556f506c005e std::_Hashtable<>::clear() @ 0x556f50780f4f doris::vectorized::VHiveTableWriter::close() @ 0x556f50780f4f doris::vectorized::VHiveTableWriter::close() @ 0x556f5072bc4f doris::vectorized::AsyncResultWriter::process_block() @ 0x556f5072bc4f doris::vectorized::AsyncResultWriter::process_block() @ 0x556f5072cd01 std::_Function_handler<>::_M_invoke() @ 0x556f5072cd01 std::_Function_handler<>::_M_invoke() @ 0x556f2b02b73d doris::ThreadPool::dispatch_thread() @ 0x556f2b02b73d doris::ThreadPool::dispatch_thread() @ 0x556f2b008d59 doris::Thread::supervise_thread() @ 0x556f2b008d59 doris::Thread::supervise_thread() @ 0x7f2c2bfb4609 start_thread @ 0x7f2c2bfb4609 start_thread @ 0x7f2c2c261133 clone @ 0x7f2c2c261133 clone @ (nil) (unknown) *** Query id: ac4f457c003d4489-b04ac56ef05b12f0 *** *** is nereids: 1 *** *** tablet id: 0 *** *** Aborted at 1713690877 (unix time) try "date -d @1713690877" if you are using GNU date *** *** Current BE git commitID: e6f4a2f *** *** SIGABRT unknown detail explain (@0x38a6) received by PID 14502 (TID 184986 OR 0x7f21d0614700) from PID 14502; stack trace: *** @ (nil) (unknown) F20240421 17:14:37.505879 185108 hdfs_util.h:65] Check failed: _ref_cnt == 0 F20240421 17:14:37.887202 185110 hdfs_util.h:65] Check failed: _ref_cnt == 0 *** Check failure stack trace: *** @ 0x556f5236d316 google::LogMessageFatal::~LogMessageFatal() @ 0x556f2830e200 doris::io::HdfsFileWriter::~HdfsFileWriter() @ 0x556f2830e21e doris::io::HdfsFileWriter::~HdfsFileWriter() @ 0x556f507893b0 doris::vectorized::VHivePartitionWriter::~VHivePartitionWriter() 0# doris::signal::(anonymous namespace)::FailureSignalHandler(int, siginfo_t*, void*) at /root/doris/be/src/common/signal_handler.h:421 ``` The root cause is When it cannot be processed in the cache (such as when the cache is full), we will create a `fs_handler`, and the life cycle of `fs_handler` is managed by caller. We have separated the hdfs writer and can create `fs_handler` separately, so `HdfsFileSystem` and `HdfsFileWriter` may be callers. In `HdfsFileSystem`, the `fs_handler` of `HdfsFileWriter` is shared, so it needs to be changed to `shared_ptr`. ### Solution Fix hdfs file writer core with `check failed: _ref_cnt == 0` in dtor of `HdfsFileWriter`. Change `fs_handler` ptr to `shared_ptr` and remove ref count operations.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Proposed changes
Issue:
The root cause is
When it cannot be processed in the cache (such as when the cache is full), we will create a
fs_handler, and the life cycle offs_handleris managed by caller. We have separated the hdfs writer and can createfs_handlerseparately, soHdfsFileSystemandHdfsFileWritermay be callers. InHdfsFileSystem, thefs_handlerofHdfsFileWriteris shared, so it needs to be changed toshared_ptr.Solution
Fix hdfs file writer core with
check failed: _ref_cnt == 0in dtor ofHdfsFileWriter.Change
fs_handlerptr toshared_ptrand remove ref count operations.Further comments
If this is a relatively large or complex change, kick off the discussion at dev@doris.apache.org by explaining why you chose the solution you did and what alternatives you considered, etc...