Skip to content

fix(cpp): never seal empty chunks and fix dangling ref in parallel tablet write #909

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
kkzi wants to merge 2 commits into
base: develop
Choose a base branch
from
+24 −8
Open

fix(cpp): never seal empty chunks and fix dangling ref in parallel tablet write #909

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
26139ba
fix(cpp): never seal empty chunks and fix dangling ref in parallel ta…
Aug 18, 2026
6363b02
style(cpp): fix spotless clang-format violations in parallel tablet w…
Aug 19, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
32 changes: 24 additions & 8 deletions cpp/src/writer/tsfile_writer.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1351,17 +1351,24 @@ int TsFileWriter::write_table(Tablet& tablet) {
common::g_thread_pool_ != nullptr) {
std::vector<std::future<int>> futures;
for (auto& ctx : device_ctxs) {
// Capture pointers by value: the submitted tasks run on pool
// threads asynchronously (after this loop returns), so
// capturing the loop variables by reference would dangle
// (all tasks would read the same/out-of-scope ctx).
auto* ctx_ptr = &ctx;
futures.push_back(common::g_thread_pool_->submit(
[&write_time_segments, &ctx]() {
return write_time_segments(ctx.tcw, ctx.segments,
ctx.initial_page_points);
[&write_time_segments, ctx_ptr]() {
return write_time_segments(
ctx_ptr->tcw, ctx_ptr->segments,
ctx_ptr->initial_page_points);
}));
for (auto& vt : ctx.value_tasks) {
auto* vt_ptr = &vt;
futures.push_back(common::g_thread_pool_->submit(
[&write_value_segments, &vt, &ctx]() {
[&write_value_segments, vt_ptr, ctx_ptr]() {
return write_value_segments(
vt.vcw, vt.col_idx, ctx.segments,
ctx.initial_page_points);
vt_ptr->vcw, vt_ptr->col_idx, ctx_ptr->segments,
ctx_ptr->initial_page_points);
}));
}
}
Expand Down Expand Up @@ -1898,7 +1905,13 @@ int TsFileWriter::flush_chunk_group_encoded(MeasurementSchemaGroup* chunk_group,
for (MeasurementSchemaMapIter ms_iter = map.begin(); ms_iter != map.end();
ms_iter++) {
MeasurementSchema* m_schema = ms_iter->second;
if (!chunk_group->is_aligned_ && m_schema->chunk_writer_ != nullptr) {
// Skip registered-but-empty columns: a measurement that was never
// written in this window would otherwise be sealed as an EMPTY chunk
// (count=0, dataSize=0). Java readers (TsFileSequenceReader self-
// check) treat such a file as crashed. Mirror the aligned branch's
// hasData() check below.
if (!chunk_group->is_aligned_ && m_schema->chunk_writer_ != nullptr &&
m_schema->chunk_writer_->hasData()) {
ChunkWriter*& chunk_writer = m_schema->chunk_writer_;
FLUSH_CHUNK_ENCODED(
chunk_writer, io_writer_, m_schema->measurement_name_,
Expand Down Expand Up @@ -1935,7 +1948,10 @@ int TsFileWriter::flush_chunk_group(MeasurementSchemaGroup* chunk_group,
for (MeasurementSchemaMapIter ms_iter = map.begin(); ms_iter != map.end();
ms_iter++) {
MeasurementSchema* m_schema = ms_iter->second;
if (!chunk_group->is_aligned_ && m_schema->chunk_writer_ != nullptr) {
// See flush_chunk_group_encoded: never seal a registered-but-empty
// column as a count=0 chunk.
if (!chunk_group->is_aligned_ && m_schema->chunk_writer_ != nullptr &&
m_schema->chunk_writer_->hasData()) {
ChunkWriter*& chunk_writer = m_schema->chunk_writer_;
FLUSH_CHUNK(chunk_writer, io_writer_, m_schema->measurement_name_,
m_schema->data_type_, m_schema->encoding_,
Expand Down
Loading