This is an automated email from the ASF dual-hosted git repository.
zouxinyi 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 d9d45198ea4 [fix](memory) Fix `SwitchResourceContext` check attach
task (#48429)
d9d45198ea4 is described below
commit d9d45198ea462682c68eab864d79094f1177ab3a
Author: Xinyi Zou <[email protected]>
AuthorDate: Mon Mar 3 18:26:29 2025 +0800
[fix](memory) Fix `SwitchResourceContext` check attach task (#48429)
### What problem does this PR solve?
Fix: #47462
`SwitchResourceContext` no longer checks whether `SCOPED_ATTACH_TASK`
has been called before, which makes `SwitchResourceContext` more
flexible.
For example, when calling the `MemTableWriter::flush_async` function, it
is correct whether the upper code attaches the task or not.
Fix:
```
F 2025-02-27 02:20:55,511 11069 thread_context.h:201] Check failed:
is_attach_task()
*** Check failure stack trace: ***
@ 0x558647d61466 google::LogMessage::SendToLog()
@ 0x558647d5deb0 google::LogMessage::Flush()
@ 0x558647d61ca9 google::LogMessageFatal::~LogMessageFatal()
@ 0x55863da71801
doris::SwitchResourceContext::SwitchResourceContext()
@ 0x55863d3b90f9 doris::MemTableWriter::flush_async()
@ 0x55863d3b0d60
doris::MemTableMemoryLimiter::_flush_active_memtables()
@ 0x55863d3b1217
doris::MemTableMemoryLimiter::flush_workload_group_memtables()
@ 0x55863da900dc
doris::WorkloadGroupMgr::flush_memtable_from_current_group_()
@ 0x55863da8d1be doris::WorkloadGroupMgr::handle_paused_queries()
@ 0x55863cb74f65 doris::Daemon::memory_maintenance_thread()
@ 0x55863dc1c631 doris::Thread::supervise_thread()
@ 0x7f60f7117ac3 (unknown)
@ 0x7f60f71a9850 (unknown)
@ (nil) (unknown)
*** Query id: 0-0 ***
*** is nereids: 0 ***
*** tablet id: 0 ***
*** Aborted at 1740594055 (unix time) try "date -d @1740594055" if you are
using GNU date ***
*** Current BE git commitID: e0d85882bd ***
*** SIGABRT unknown detail explain (@0x22fc) received by PID 8956 (TID
11069 OR 0x7f5d09c04640) from PID 8956; stack trace: ***
0# doris::signal::(anonymous namespace)::FailureSignalHandler(int,
siginfo_t*, void*) at
/home/zcp/repo_center/doris_master/doris/be/src/common/signal_handler.h:421
1# 0x00007F60F70C5520 in /lib/x86_64-linux-gnu/libc.so.6
2# pthread_kill at ./nptl/pthread_kill.c:89
3# raise at ../sysdeps/posix/raise.c:27
4# abort at ./stdlib/abort.c:81
5# 0x0000558647D6BD3D in
/mnt/hdd01/ci/doris-deploy-master-local/be/lib/doris_be
6# 0x0000558647D5E37A in
/mnt/hdd01/ci/doris-deploy-master-local/be/lib/doris_be
7# google::LogMessage::SendToLog() in
/mnt/hdd01/ci/doris-deploy-master-local/be/lib/doris_be
8# google::LogMessage::Flush() in
/mnt/hdd01/ci/doris-deploy-master-local/be/lib/doris_be
9# google::LogMessageFatal::~LogMessageFatal() in
/mnt/hdd01/ci/doris-deploy-master-local/be/lib/doris_be
10#
doris::SwitchResourceContext::SwitchResourceContext(std::shared_ptr<doris::ResourceContext>
const&) in /mnt/hdd01/ci/doris-deploy-master-local/be/lib/doris_be
11# doris::MemTableWriter::flush_async() at
/home/zcp/repo_center/doris_master/doris/be/src/olap/memtable_writer.cpp:159
12# doris::MemTableMemoryLimiter::_flush_active_memtables(unsigned long,
long) at
/home/zcp/repo_center/doris_master/doris/be/src/olap/memtable_memory_limiter.cpp:262
13# doris::MemTableMemoryLimiter::flush_workload_group_memtables(unsigned
long, long) in /mnt/hdd01/ci/doris-deploy-master-local/be/lib/doris_be
14#
doris::WorkloadGroupMgr::flush_memtable_from_current_group_(std::shared_ptr<doris::WorkloadGroup>,
long) at
/home/zcp/repo_center/doris_master/doris/be/src/runtime/workload_group/workload_group_manager.cpp:568
15# doris::WorkloadGroupMgr::handle_paused_queries() in
/mnt/hdd01/ci/doris-deploy-master-local/be/lib/doris_be
16# doris::Daemon::memory_maintenance_thread() at
/home/zcp/repo_center/doris_master/doris/be/src/common/daemon.cpp:329
17# doris::Thread::supervise_thread(void*) at
/home/zcp/repo_center/doris_master/doris/be/src/util/thread.cpp:499
18# start_thread at ./nptl/pthread_create.c:442
19# 0x00007F60F71A9850 at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:83
```
---
be/src/olap/memtable_writer.cpp | 3 ++-
be/src/runtime/load_channel.cpp | 1 +
be/src/runtime/load_channel.h | 2 --
be/src/runtime/load_channel_mgr.cpp | 1 -
be/src/runtime/thread_context.cpp | 1 -
5 files changed, 3 insertions(+), 5 deletions(-)
diff --git a/be/src/olap/memtable_writer.cpp b/be/src/olap/memtable_writer.cpp
index 1a0e1e99b69..25efc82612b 100644
--- a/be/src/olap/memtable_writer.cpp
+++ b/be/src/olap/memtable_writer.cpp
@@ -152,9 +152,10 @@ Status MemTableWriter::_flush_memtable_async() {
Status MemTableWriter::flush_async() {
std::lock_guard<std::mutex> l(_lock);
- // Two calling paths:
+ // Three calling paths:
// 1. call by local, from `VTabletWriterV2::_write_memtable`.
// 2. call by remote, from `LoadChannelMgr::_get_load_channel`.
+ // 3. call by daemon thread, from `handle_paused_queries` ->
`flush_workload_group_memtables`.
SCOPED_SWITCH_RESOURCE_CONTEXT(_resource_ctx);
if (!_is_init || _is_closed) {
// This writer is uninitialized or closed before flushing, do nothing.
diff --git a/be/src/runtime/load_channel.cpp b/be/src/runtime/load_channel.cpp
index ce10666d84c..d745132e589 100644
--- a/be/src/runtime/load_channel.cpp
+++ b/be/src/runtime/load_channel.cpp
@@ -174,6 +174,7 @@ Status LoadChannel::add_batch(const
PTabletWriterAddBlockRequest& request,
PTabletWriterAddBlockResult* response) {
SCOPED_TIMER(_add_batch_timer);
COUNTER_UPDATE(_add_batch_times, 1);
+ SCOPED_ATTACH_TASK(_resource_ctx);
int64_t index_id = request.index_id();
// 1. get tablets channel
std::shared_ptr<BaseTabletsChannel> channel;
diff --git a/be/src/runtime/load_channel.h b/be/src/runtime/load_channel.h
index 8db5ea50993..31164222a95 100644
--- a/be/src/runtime/load_channel.h
+++ b/be/src/runtime/load_channel.h
@@ -69,8 +69,6 @@ public:
bool is_high_priority() const { return _is_high_priority; }
- std::shared_ptr<ResourceContext> resource_ctx() const { return
_resource_ctx; }
-
WorkloadGroupPtr workload_group() const { return
_resource_ctx->workload_group(); }
RuntimeProfile::Counter* get_mgr_add_batch_timer() { return
_mgr_add_batch_timer; }
diff --git a/be/src/runtime/load_channel_mgr.cpp
b/be/src/runtime/load_channel_mgr.cpp
index 0bf1bf18670..55db6564488 100644
--- a/be/src/runtime/load_channel_mgr.cpp
+++ b/be/src/runtime/load_channel_mgr.cpp
@@ -144,7 +144,6 @@ Status LoadChannelMgr::add_batch(const
PTabletWriterAddBlockRequest& request,
return status;
}
SCOPED_TIMER(channel->get_mgr_add_batch_timer());
- SCOPED_ATTACH_TASK(channel->resource_ctx());
if (!channel->is_high_priority()) {
// 2. check if mem consumption exceed limit
diff --git a/be/src/runtime/thread_context.cpp
b/be/src/runtime/thread_context.cpp
index 266515ae2a6..7db5c06d1a9 100644
--- a/be/src/runtime/thread_context.cpp
+++ b/be/src/runtime/thread_context.cpp
@@ -57,7 +57,6 @@ AttachTask::~AttachTask() {
SwitchResourceContext::SwitchResourceContext(const
std::shared_ptr<ResourceContext>& rc) {
DCHECK(rc != nullptr);
- DCHECK(thread_context()->is_attach_task());
doris::ThreadLocalHandle::create_thread_local_if_not_exits();
if (rc != thread_context()->resource_ctx()) {
signal::set_signal_task_id(rc->task_controller()->task_id());
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
