[fix](load) fix quorum success invalid in move-memtable-on-sink load path#60681
Merged
dataroaring merged 1 commit intoapache:masterfrom Feb 25, 2026
Merged
[fix](load) fix quorum success invalid in move-memtable-on-sink load path#60681dataroaring merged 1 commit intoapache:masterfrom
dataroaring merged 1 commit intoapache:masterfrom
Conversation
Contributor
|
Thank you for your contribution to Apache Doris. Please clearly describe your PR:
|
Contributor
Author
|
run buildall |
TPC-H: Total hot run time: 30175 ms |
ClickBench: Total hot run time: 28.36 s |
Contributor
BE UT Coverage ReportIncrement line coverage Increment coverage report
|
Contributor
BE Regression && UT Coverage ReportIncrement line coverage Increment coverage report
|
sollhui
force-pushed
the
fix_quorum_sucess
branch
from
6cbd3bf to
4904e66
Compare
sollhui
changed the title
Contributor
Author
|
run buildall |
TPC-H: Total hot run time: 30899 ms |
ClickBench: Total hot run time: 28.27 s |
Contributor
Author
|
run buildall |
TPC-H: Total hot run time: 29037 ms |
TPC-DS: Total hot run time: 183418 ms |
BE UT Coverage ReportIncrement line coverage Increment coverage report
|
Contributor
BE Regression && UT Coverage ReportIncrement line coverage Increment coverage report
|
Contributor
BE Regression && UT Coverage ReportIncrement line coverage Increment coverage report
|
1 similar comment
Contributor
BE Regression && UT Coverage ReportIncrement line coverage Increment coverage report
|
Contributor
BE Regression && UT Coverage ReportIncrement line coverage Increment coverage report
|
Contributor
BE Regression && UT Coverage ReportIncrement line coverage Increment coverage report
|
Contributor
Author
|
run buildall |
TPC-H: Total hot run time: 28791 ms |
TPC-DS: Total hot run time: 184521 ms |
BE UT Coverage ReportIncrement line coverage Increment coverage report
|
Contributor
BE Regression && UT Coverage ReportIncrement line coverage Increment coverage report
|
Contributor
BE Regression && UT Coverage ReportIncrement line coverage Increment coverage report
|
Contributor
BE Regression && UT Coverage ReportIncrement line coverage Increment coverage report
|
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. |
github-actions bot
pushed a commit
that referenced
this pull request
Feb 25, 2026
…path (#60681) ## Description In `VTabletWriterV2`, the `_quorum_success()` function always returns `false` even when all streams have finished successfully, making the quorum success write feature effectively non-functional. ## Root Cause In both `_build_tablet_node_mapping()` and `_incremental_open_streams()`, `_tablets_by_node[node].emplace(tablet_id)` is guarded by the `known_indexes` check: ```cpp if (known_indexes.contains(index.index_id)) [[likely]] { continue; } _indexes_from_node[node].emplace_back(tablet); _tablets_by_node[node].emplace(tablet_id); known_indexes.insert(index.index_id); ``` The known_indexes set is shared across all partitions, tablets, and nodes. Once an index_id is inserted after processing the first tablet's first node, all subsequent tablets (and all other nodes of the same tablet) with the same index_id skip the _tablets_by_node update. For example, with 1 index, 3 tablets [T1, T2, T3], and 3 replicas [N1, N2, N3]: Only _tablets_by_node[N1] = {T1} gets populated T1 on N2/N3, and T2/T3 on all nodes are skipped This causes _quorum_success() to compute finished_tablets_replica incorrectly: finished_tablets_replica[T1] = 1 (only counted from N1) finished_tablets_replica[T2] = 0, finished_tablets_replica[T3] = 0 With a quorum requirement of 2 (for 3 replicas), the check always fails. The known_indexes optimization was intended only for _indexes_from_node (to avoid sending duplicate schema info per index), but it incorrectly also blocked the _tablets_by_node population. Note: vtablet_writer.cpp does NOT have this issue — its _tablets_by_channel is populated without the known_indexes guard. ## Fix Move _tablets_by_node[node].emplace(tablet_id) before the known_indexes check in both _build_tablet_node_mapping() and _incremental_open_streams(), so that every tablet on every node is correctly recorded.
ybtsdst
pushed a commit
to ybtsdst/doris
that referenced
this pull request
Feb 27, 2026
…n-sink load path apache#60681 (apache#60820) Cherry-picked from apache#60681 Co-authored-by: hui lai <laihui@selectdb.com>
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
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.
Description
In
VTabletWriterV2, the_quorum_success()function always returnsfalseeven whenall streams have finished successfully, making the quorum success write feature effectively
non-functional.
Root Cause
In both
_build_tablet_node_mapping()and_incremental_open_streams(),_tablets_by_node[node].emplace(tablet_id)is guarded by theknown_indexescheck:The known_indexes set is shared across all partitions, tablets, and nodes. Once an
index_id is inserted after processing the first tablet's first node, all subsequent
tablets (and all other nodes of the same tablet) with the same index_id skip the
_tablets_by_node update.
For example, with 1 index, 3 tablets [T1, T2, T3], and 3 replicas [N1, N2, N3]:
Only _tablets_by_node[N1] = {T1} gets populated
T1 on N2/N3, and T2/T3 on all nodes are skipped
This causes _quorum_success() to compute finished_tablets_replica incorrectly:
finished_tablets_replica[T1] = 1 (only counted from N1)
finished_tablets_replica[T2] = 0, finished_tablets_replica[T3] = 0
With a quorum requirement of 2 (for 3 replicas), the check always fails.
The known_indexes optimization was intended only for _indexes_from_node (to avoid
sending duplicate schema info per index), but it incorrectly also blocked the
_tablets_by_node population.
Note: vtablet_writer.cpp does NOT have this issue — its _tablets_by_channel is
populated without the known_indexes guard.
Fix
Move _tablets_by_node[node].emplace(tablet_id) before the known_indexes check in
both _build_tablet_node_mapping() and _incremental_open_streams(), so that every
tablet on every node is correctly recorded.