This is an automated email from the ASF dual-hosted git repository.
yiguolei pushed a commit to branch branch-4.0
in repository https://gitbox.apache.org/repos/asf/doris.git
The following commit(s) were added to refs/heads/branch-4.0 by this push:
new 1bcfd51933b branch-4.0: [opt](memory) implement freeing memory column
by column for partial update (#58275) (#58843)
1bcfd51933b is described below
commit 1bcfd51933b1e924e74d3f9da969ad32b964f527
Author: hui lai <[email protected]>
AuthorDate: Mon Dec 29 10:01:44 2025 +0800
branch-4.0: [opt](memory) implement freeing memory column by column for
partial update (#58275) (#58843)
pick #58275
Now, pages of column writer will be released after write all memtable
block in the scenario of partial updates, the process is as follows:
1. create all column_writer
2. append all column_writer (significant amount of memory expansion)
3. finish and free memory. This implement will cause a significant
amount of memory expansion when flushing memtable(The larger the number
of columns, the more obvious it is), and make load could easily
cancelled by memory manager when table column too wide(such as 5000
column).
This pr implement freeing memory column by column for partial update to
solve the problem, the process is as follows:
1. create all column_writer
2. append column_writer and free memory column by column. Test using
machines with specifications of 16c and 64G with 5000 column table:
- Before this pr, `VerticalSegmentWriter::_create_column_writer`
occupies nearly 20% of the memory.
- After fix, ` VerticalSegmentWriter::_create_column_writer` hardly
occupies memory.
### What problem does this PR solve?
Issue Number: close #xxx
Related PR: #xxx
Problem Summary:
### Release note
None
### Check List (For Author)
- Test <!-- At least one of them must be included. -->
- [ ] Regression test
- [ ] Unit Test
- [ ] Manual test (add detailed scripts or steps below)
- [ ] No need to test or manual test. Explain why:
- [ ] This is a refactor/code format and no logic has been changed.
- [ ] Previous test can cover this change.
- [ ] No code files have been changed.
- [ ] Other reason <!-- Add your reason? -->
- Behavior changed:
- [ ] No.
- [ ] Yes. <!-- Explain the behavior change -->
- Does this need documentation?
- [ ] No.
- [ ] Yes. <!-- Add document PR link here. eg:
https://github.com/apache/doris-website/pull/1214 -->
### Check List (For Reviewer who merge this PR)
- [ ] Confirm the release note
- [ ] Confirm test cases
- [ ] Confirm document
- [ ] Add branch pick label <!-- Add branch pick label that this PR
should merge into -->
---
.../rowset/segment_v2/vertical_segment_writer.cpp | 38 ++++++++++------------
.../rowset/segment_v2/vertical_segment_writer.h | 2 ++
2 files changed, 20 insertions(+), 20 deletions(-)
diff --git a/be/src/olap/rowset/segment_v2/vertical_segment_writer.cpp
b/be/src/olap/rowset/segment_v2/vertical_segment_writer.cpp
index 975a9970252..f9cc984a370 100644
--- a/be/src/olap/rowset/segment_v2/vertical_segment_writer.cpp
+++ b/be/src/olap/rowset/segment_v2/vertical_segment_writer.cpp
@@ -451,6 +451,19 @@ Status VerticalSegmentWriter::_probe_key_for_mow(
return Status::OK();
}
+Status VerticalSegmentWriter::_finalize_column_writer_and_update_meta(size_t
cid) {
+ RETURN_IF_ERROR(_column_writers[cid]->finish());
+ RETURN_IF_ERROR(_column_writers[cid]->write_data());
+
+ auto* column_meta = _column_writers[cid]->get_column_meta();
+ column_meta->set_compressed_data_bytes(
+ _column_writers[cid]->get_total_compressed_data_pages_bytes());
+ column_meta->set_uncompressed_data_bytes(
+ _column_writers[cid]->get_total_uncompressed_data_pages_bytes());
+
column_meta->set_raw_data_bytes(_column_writers[cid]->get_raw_data_bytes());
+ return Status::OK();
+}
+
Status VerticalSegmentWriter::_partial_update_preconditions_check(size_t
row_pos,
bool
is_flexible_update) {
if (!_is_mow()) {
@@ -540,6 +553,7 @@ Status
VerticalSegmentWriter::_append_block_with_partial_content(RowsInBlock& da
}
RETURN_IF_ERROR(_column_writers[cid]->append(column->get_nullmap(),
column->get_data(),
data.num_rows));
+ RETURN_IF_ERROR(_finalize_column_writer_and_update_meta(cid));
}
bool has_default_or_nullable = false;
@@ -629,6 +643,7 @@ Status
VerticalSegmentWriter::_append_block_with_partial_content(RowsInBlock& da
}
RETURN_IF_ERROR(_column_writers[cid]->append(column->get_nullmap(),
column->get_data(),
data.num_rows));
+ RETURN_IF_ERROR(_finalize_column_writer_and_update_meta(cid));
}
_num_rows_updated += stats.num_rows_updated;
@@ -726,6 +741,7 @@ Status
VerticalSegmentWriter::_append_block_with_flexible_partial_content(
RETURN_IF_ERROR(_column_writers[cid]->append(column->get_nullmap(),
column->get_data(),
data.num_rows));
DCHECK(_column_writers[cid]->get_next_rowid() == _num_rows_written +
data.num_rows);
+ RETURN_IF_ERROR(_finalize_column_writer_and_update_meta(cid));
}
// 5. genreate read plan
@@ -771,6 +787,7 @@ Status
VerticalSegmentWriter::_append_block_with_flexible_partial_content(
RETURN_IF_ERROR(_column_writers[cid]->append(column->get_nullmap(),
column->get_data(),
data.num_rows));
DCHECK(_column_writers[cid]->get_next_rowid() == _num_rows_written +
data.num_rows);
+ RETURN_IF_ERROR(_finalize_column_writer_and_update_meta(cid));
}
_num_rows_updated += stats.num_rows_updated;
@@ -932,17 +949,6 @@ Status VerticalSegmentWriter::write_batch() {
RETURN_IF_ERROR(_append_block_with_partial_content(data,
full_block));
}
}
- for (auto& column_writer : _column_writers) {
- RETURN_IF_ERROR(column_writer->finish());
- RETURN_IF_ERROR(column_writer->write_data());
-
- auto* column_meta = column_writer->get_column_meta();
- column_meta->set_compressed_data_bytes(
- column_writer->get_total_compressed_data_pages_bytes());
- column_meta->set_uncompressed_data_bytes(
- column_writer->get_total_uncompressed_data_pages_bytes());
-
column_meta->set_raw_data_bytes(column_writer->get_raw_data_bytes());
- }
return Status::OK();
}
// Row column should be filled here when it's a directly write from
memtable
@@ -992,15 +998,7 @@ Status VerticalSegmentWriter::write_batch() {
return Status::Error<DISK_REACH_CAPACITY_LIMIT>("disk {} exceed
capacity limit.",
_data_dir->path_hash());
}
- RETURN_IF_ERROR(_column_writers[cid]->finish());
- RETURN_IF_ERROR(_column_writers[cid]->write_data());
-
- auto* column_meta = _column_writers[cid]->get_column_meta();
- column_meta->set_compressed_data_bytes(
- _column_writers[cid]->get_total_compressed_data_pages_bytes());
- column_meta->set_uncompressed_data_bytes(
-
_column_writers[cid]->get_total_uncompressed_data_pages_bytes());
-
column_meta->set_raw_data_bytes(_column_writers[cid]->get_raw_data_bytes());
+ RETURN_IF_ERROR(_finalize_column_writer_and_update_meta(cid));
}
for (auto& data : _batched_blocks) {
diff --git a/be/src/olap/rowset/segment_v2/vertical_segment_writer.h
b/be/src/olap/rowset/segment_v2/vertical_segment_writer.h
index 7bf795a7d62..9528f12f5c7 100644
--- a/be/src/olap/rowset/segment_v2/vertical_segment_writer.h
+++ b/be/src/olap/rowset/segment_v2/vertical_segment_writer.h
@@ -198,6 +198,8 @@ private:
vectorized::IOlapColumnDataAccessor* seq_column, size_t num_rows,
bool need_sort);
Status
_generate_short_key_index(std::vector<vectorized::IOlapColumnDataAccessor*>&
key_columns,
size_t num_rows, const
std::vector<size_t>& short_key_pos);
+ Status _finalize_column_writer_and_update_meta(size_t cid);
+
bool _is_mow();
bool _is_mow_with_cluster_key();
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
