This is an automated email from the ASF dual-hosted git repository.
yiguolei pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/doris.git
The following commit(s) were added to refs/heads/master by this push:
new d0165e47b08 [improvement](be) Add instance ID to hash join profiles
(#66097)
d0165e47b08 is described below
commit d0165e47b08d5c3c56cbfc672760262a4e6a0cb1
Author: Jerry Hu <[email protected]>
AuthorDate: Wed Aug 5 09:05:48 2026 +0800
[improvement](be) Add instance ID to hash join profiles (#66097)
### What problem does this PR solve?
Issue Number: None
Related PR: #65761
Problem Summary:
This forward-ports #65761 from `opt_perf_4.1` to `master`.
Shuffle hash join build and probe operators run in separate pipelines,
so their per-instance profiles are difficult to correlate. Add the
fragment instance ID to both hash join build and probe custom profiles.
Profiles with the same join node ID and `InstanceID` now identify the
corresponding build/probe pair without relying on process-local hash
table addresses.
### Release note
None
### Check List (For Author)
- Test
- [ ] Regression test
- [ ] Unit Test
- [ ] Manual test
- [x] No need to test or manual test. The change only adds profile
metadata.
- `build-support/clang-format.sh` (passed)
- `build-support/check-format.sh` (passed)
- `git diff --check origin/master...HEAD` (passed)
- `BUILD_TYPE=ASAN ./build.sh --be` (attempted; CMake configured
successfully, but Ninja was blocked before compilation because this
worktree lacks `thirdparty/installed/lib64/liblance_c.a`)
- `thirdparty/build-thirdparty.sh lance_c` (attempted; blocked because
the required Rust/Cargo 1.91 toolchain is unavailable)
- `build-support/run-clang-tidy.sh --base origin/master --build-dir
be/build_ASAN` (attempted; blocked by the existing `core/types.h`
unmatched `NOLINTEND` and missing system `stddef.h`; no diagnostic
identified on the added lines)
- [ ] This is a refactor/code format and no logic has been changed.
- Behavior changed:
- [ ] No.
- [x] Yes. Hash join build and probe profiles include `InstanceID`.
- Does this need documentation?
- [x] No.
- [ ] Yes.
Co-authored-by: HappenLee <[email protected]>
---
be/src/exec/operator/hashjoin_build_sink.cpp | 1 +
be/src/exec/operator/hashjoin_probe_operator.cpp | 2 ++
2 files changed, 3 insertions(+)
diff --git a/be/src/exec/operator/hashjoin_build_sink.cpp
b/be/src/exec/operator/hashjoin_build_sink.cpp
index ebd90a462ec..55716412701 100644
--- a/be/src/exec/operator/hashjoin_build_sink.cpp
+++ b/be/src/exec/operator/hashjoin_build_sink.cpp
@@ -56,6 +56,7 @@ Status HashJoinBuildSinkLocalState::init(RuntimeState* state,
LocalSinkStateInfo
_task_idx = info.task_idx;
auto& p = _parent->cast<HashJoinBuildSinkOperatorX>();
_shared_state->join_op_variants = p._join_op_variants;
+ custom_profile()->add_info_string("InstanceID",
print_id(state->fragment_instance_id()));
_build_expr_ctxs.resize(p._build_expr_ctxs.size());
for (size_t i = 0; i < _build_expr_ctxs.size(); i++) {
diff --git a/be/src/exec/operator/hashjoin_probe_operator.cpp
b/be/src/exec/operator/hashjoin_probe_operator.cpp
index 80543c55e4e..44d9e806672 100644
--- a/be/src/exec/operator/hashjoin_probe_operator.cpp
+++ b/be/src/exec/operator/hashjoin_probe_operator.cpp
@@ -30,6 +30,7 @@
#include "core/data_type/data_type_nullable.h"
#include "exec/operator/operator.h"
#include "runtime/descriptors.h"
+#include "util/uid_util.h"
namespace doris {
HashJoinProbeLocalState::HashJoinProbeLocalState(RuntimeState* state,
OperatorXBase* parent)
@@ -42,6 +43,7 @@ Status HashJoinProbeLocalState::init(RuntimeState* state,
LocalStateInfo& info)
SCOPED_TIMER(_init_timer);
_task_idx = info.task_idx;
auto& p = _parent->cast<HashJoinProbeOperatorX>();
+ custom_profile()->add_info_string("InstanceID",
print_id(state->fragment_instance_id()));
_probe_expr_ctxs.resize(p._probe_expr_ctxs.size());
for (size_t i = 0; i < _probe_expr_ctxs.size(); i++) {
RETURN_IF_ERROR(p._probe_expr_ctxs[i]->clone(state,
_probe_expr_ctxs[i]));
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
