This is an automated email from the ASF dual-hosted git repository.
wangbo 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 6637f9c15f9 Add enable_cgroup_cpu_soft_limit (#26510)
6637f9c15f9 is described below
commit 6637f9c15f9346abbd00feb41eca9624a7bc1c1e
Author: wangbo <wan...@apache.org>
AuthorDate: Wed Nov 8 15:52:13 2023 +0800
Add enable_cgroup_cpu_soft_limit (#26510)
---
be/src/agent/cgroup_cpu_ctl.cpp | 24 ++++++++++++++++++++++--
be/src/agent/cgroup_cpu_ctl.h | 22 +++++++++++++++++++---
be/src/common/config.cpp | 1 +
be/src/common/config.h | 2 ++
be/src/runtime/fragment_mgr.cpp | 21 +++++++++++++++++----
be/src/runtime/task_group/task_group_manager.cpp | 12 ++++++++++--
be/src/runtime/task_group/task_group_manager.h | 3 ++-
7 files changed, 73 insertions(+), 12 deletions(-)
diff --git a/be/src/agent/cgroup_cpu_ctl.cpp b/be/src/agent/cgroup_cpu_ctl.cpp
index ee24b34e167..d16a32b7be5 100644
--- a/be/src/agent/cgroup_cpu_ctl.cpp
+++ b/be/src/agent/cgroup_cpu_ctl.cpp
@@ -54,6 +54,19 @@ void CgroupCpuCtl::update_cpu_hard_limit(int cpu_hard_limit)
{
}
}
+void CgroupCpuCtl::update_cpu_soft_limit(int cpu_shares) {
+ if (!_init_succ) {
+ return;
+ }
+ std::lock_guard<std::shared_mutex> w_lock(_lock_mutex);
+ if (_cpu_shares != cpu_shares) {
+ Status ret = modify_cg_cpu_soft_limit_no_lock(cpu_shares);
+ if (ret.ok()) {
+ _cpu_shares = cpu_shares;
+ }
+ }
+}
+
Status CgroupCpuCtl::write_cg_sys_file(std::string file_path, int value,
std::string msg,
bool is_append) {
int fd = open(file_path.c_str(), is_append ? O_RDWR | O_APPEND : O_RDWR);
@@ -97,9 +110,11 @@ Status CgroupV1CpuCtl::init() {
}
}
- // quota path
+ // quota file
_cgroup_v1_cpu_tg_quota_file = _cgroup_v1_cpu_tg_path +
"/cpu.cfs_quota_us";
- // task path
+ // cpu.shares file
+ _cgroup_v1_cpu_tg_shares_file = _cgroup_v1_cpu_tg_path + "/cpu.shares";
+ // task file
_cgroup_v1_cpu_tg_task_file = _cgroup_v1_cpu_tg_path + "/tasks";
LOG(INFO) << "cgroup v1 cpu path init success"
<< ", query tg path=" << _cgroup_v1_cpu_tg_path
@@ -110,6 +125,11 @@ Status CgroupV1CpuCtl::init() {
return Status::OK();
}
+Status CgroupV1CpuCtl::modify_cg_cpu_soft_limit_no_lock(int cpu_shares) {
+ std::string msg = "modify cpu shares to " + std::to_string(cpu_shares);
+ return CgroupCpuCtl::write_cg_sys_file(_cgroup_v1_cpu_tg_shares_file,
cpu_shares, msg, false);
+}
+
Status CgroupV1CpuCtl::modify_cg_cpu_hard_limit_no_lock(int cpu_hard_limit) {
int val = _cpu_cfs_period_us * _cpu_core_num * cpu_hard_limit / 100;
std::string msg = "modify cpu quota value to " + std::to_string(val);
diff --git a/be/src/agent/cgroup_cpu_ctl.h b/be/src/agent/cgroup_cpu_ctl.h
index c3a30660147..b98e268da09 100644
--- a/be/src/agent/cgroup_cpu_ctl.h
+++ b/be/src/agent/cgroup_cpu_ctl.h
@@ -28,6 +28,12 @@
namespace doris {
+// cgroup cpu.cfs_quota_us default value, it means disable cpu hard limit
+const static int CPU_HARD_LIMIT_DEFAULT_VALUE = -1;
+
+// cgroup cpu.shares default value
+const static uint64_t CPU_SOFT_LIMIT_DEFAULT_VALUE = 1024;
+
class CgroupCpuCtl {
public:
virtual ~CgroupCpuCtl() = default;
@@ -35,15 +41,19 @@ public:
virtual Status init();
- virtual Status modify_cg_cpu_hard_limit_no_lock(int cpu_hard_limit) = 0;
-
virtual Status add_thread_to_cgroup() = 0;
void update_cpu_hard_limit(int cpu_hard_limit);
+ void update_cpu_soft_limit(int cpu_shares);
+
protected:
Status write_cg_sys_file(std::string file_path, int value, std::string
msg, bool is_append);
+ virtual Status modify_cg_cpu_hard_limit_no_lock(int cpu_hard_limit) = 0;
+
+ virtual Status modify_cg_cpu_soft_limit_no_lock(int cpu_shares) = 0;
+
std::string _doris_cgroup_cpu_path;
uint64_t _cpu_core_num = CpuInfo::num_cores();
uint64_t _cpu_cfs_period_us = 100000;
@@ -51,6 +61,7 @@ protected:
std::shared_mutex _lock_mutex;
bool _init_succ = false;
uint64_t _tg_id; // workload group id
+ uint64_t _cpu_shares = 0;
};
/*
@@ -73,20 +84,25 @@ protected:
6 workload group quota file:
/sys/fs/cgroup/cpu/{doris_home}/query/{workload group
id}/cpu.cfs_quota_us
- 7 workload group tasks file:
+ 7 workload group tasks file:
/sys/fs/cgroup/cpu/{doris_home}/query/{workload group id}/tasks
+
+ 8 workload group cpu.shares file:
+ /sys/fs/cgroup/cpu/{doris_home}/query/{workload group id}/cpu.shares
*/
class CgroupV1CpuCtl : public CgroupCpuCtl {
public:
CgroupV1CpuCtl(uint64_t tg_id) : CgroupCpuCtl(tg_id) {}
Status init() override;
Status modify_cg_cpu_hard_limit_no_lock(int cpu_hard_limit) override;
+ Status modify_cg_cpu_soft_limit_no_lock(int cpu_shares) override;
Status add_thread_to_cgroup() override;
private:
std::string _cgroup_v1_cpu_query_path;
std::string _cgroup_v1_cpu_tg_path; // workload group path
std::string _cgroup_v1_cpu_tg_quota_file;
+ std::string _cgroup_v1_cpu_tg_shares_file;
std::string _cgroup_v1_cpu_tg_task_file;
};
diff --git a/be/src/common/config.cpp b/be/src/common/config.cpp
index e858a5b6acb..17f69c3d360 100644
--- a/be/src/common/config.cpp
+++ b/be/src/common/config.cpp
@@ -1113,6 +1113,7 @@ DEFINE_Bool(enable_flush_file_cache_async, "true");
// cgroup
DEFINE_String(doris_cgroup_cpu_path, "");
+DEFINE_Bool(enable_cgroup_cpu_soft_limit, "false");
DEFINE_Bool(ignore_always_true_predicate_for_segment, "true");
diff --git a/be/src/common/config.h b/be/src/common/config.h
index 493f9342aa7..cfcd09c1984 100644
--- a/be/src/common/config.h
+++ b/be/src/common/config.h
@@ -1183,6 +1183,8 @@ DECLARE_mBool(exit_on_exception);
// cgroup
DECLARE_String(doris_cgroup_cpu_path);
+DECLARE_Bool(enable_cgroup_cpu_soft_limit);
+
// This config controls whether the s3 file writer would flush cache
asynchronously
DECLARE_Bool(enable_flush_file_cache_async);
diff --git a/be/src/runtime/fragment_mgr.cpp b/be/src/runtime/fragment_mgr.cpp
index 83b57c28fd7..266c652ca7f 100644
--- a/be/src/runtime/fragment_mgr.cpp
+++ b/be/src/runtime/fragment_mgr.cpp
@@ -671,18 +671,31 @@ Status FragmentMgr::_get_query_ctx(const Params& params,
TUniqueId query_id, boo
LOG(INFO) << "Query/load id: " <<
print_id(query_ctx->query_id())
<< " use task group: " << tg->debug_string()
<< " cpu_hard_limit: " <<
task_group_info.cpu_hard_limit
- << " cpu_share:" << task_group_info.cpu_share;
+ << " cpu_share:" << task_group_info.cpu_share
+ << " enable cgroup soft cpu:" <<
config::enable_cgroup_cpu_soft_limit;
if (task_group_info.cpu_hard_limit > 0) {
Status ret =
_exec_env->task_group_manager()->create_and_get_task_scheduler(
- tg_id, tg_name,
task_group_info.cpu_hard_limit, _exec_env,
- query_ctx.get());
+ tg_id, tg_name, task_group_info.cpu_hard_limit,
+ task_group_info.cpu_share, _exec_env,
query_ctx.get());
if (!ret.ok()) {
LOG(INFO) << "workload group init failed "
<< ", name=" << tg_name << ", id=" <<
tg_id
<< ", reason=" << ret.to_string();
}
} else {
- query_ctx->set_task_group(tg);
+ if (!config::enable_cgroup_cpu_soft_limit) {
+ query_ctx->set_task_group(tg);
+ } else {
+ Status ret =
+
_exec_env->task_group_manager()->create_and_get_task_scheduler(
+ tg_id, tg_name,
task_group_info.cpu_hard_limit,
+ task_group_info.cpu_share,
_exec_env, query_ctx.get());
+ if (!ret.ok()) {
+ LOG(INFO) << "workload group cpu soft limit
init failed "
+ << ", name=" << tg_name << ", id="
<< tg_id
+ << ", reason=" << ret.to_string();
+ }
+ }
}
}
} else {
diff --git a/be/src/runtime/task_group/task_group_manager.cpp
b/be/src/runtime/task_group/task_group_manager.cpp
index b3c24fa96e7..fb940069787 100644
--- a/be/src/runtime/task_group/task_group_manager.cpp
+++ b/be/src/runtime/task_group/task_group_manager.cpp
@@ -60,7 +60,8 @@ void TaskGroupManager::get_resource_groups(const
std::function<bool(const TaskGr
}
Status TaskGroupManager::create_and_get_task_scheduler(uint64_t tg_id,
std::string tg_name,
- int cpu_hard_limit,
ExecEnv* exec_env,
+ int cpu_hard_limit, int
cpu_shares,
+ ExecEnv* exec_env,
QueryContext*
query_ctx_ptr) {
std::lock_guard<std::mutex> lock(_task_scheduler_lock);
// step 1: init cgroup cpu controller
@@ -117,7 +118,14 @@ Status
TaskGroupManager::create_and_get_task_scheduler(uint64_t tg_id, std::stri
query_ctx_ptr->set_scan_task_scheduler(scan_task_sche);
// step 5 update cgroup cpu if needed
- _cgroup_ctl_map.at(tg_id)->update_cpu_hard_limit(cpu_hard_limit);
+ if (cpu_hard_limit > 0) {
+ _cgroup_ctl_map.at(tg_id)->update_cpu_hard_limit(cpu_hard_limit);
+
_cgroup_ctl_map.at(tg_id)->update_cpu_soft_limit(CPU_SOFT_LIMIT_DEFAULT_VALUE);
+ } else {
+ _cgroup_ctl_map.at(tg_id)->update_cpu_soft_limit(cpu_shares);
+ _cgroup_ctl_map.at(tg_id)->update_cpu_hard_limit(
+ CPU_HARD_LIMIT_DEFAULT_VALUE); // disable cpu hard limit
+ }
return Status::OK();
}
diff --git a/be/src/runtime/task_group/task_group_manager.h
b/be/src/runtime/task_group/task_group_manager.h
index ae501e93f3e..cf44f535440 100644
--- a/be/src/runtime/task_group/task_group_manager.h
+++ b/be/src/runtime/task_group/task_group_manager.h
@@ -52,7 +52,8 @@ public:
std::vector<TaskGroupPtr>* task_groups);
Status create_and_get_task_scheduler(uint64_t wg_id, std::string wg_name,
int cpu_hard_limit,
- ExecEnv* exec_env, QueryContext*
query_ctx_ptr);
+ int cpu_shares, ExecEnv* exec_env,
+ QueryContext* query_ctx_ptr);
void delete_task_group_by_ids(std::set<uint64_t> id_set);
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@doris.apache.org
For additional commands, e-mail: commits-h...@doris.apache.org
