!16632 added cluster metadata

From: @anancds
Reviewed-by: @cristoval
Signed-off-by:

mindspore/ccsrc/ps/core/abstract_node.cc

@@ -294,7 +294,7 @@ void AbstractNode::StartHeartbeatTimer(const std::shared_ptr<TcpClient> &client)
       } else {
         UpdateSchedulerTime();
       }
-      std::this_thread::sleep_for(std::chrono::seconds(ClusterMetadata::instance()->heartbeat_interval()));
+      std::this_thread::sleep_for(std::chrono::seconds(PSContext::instance()->cluster_config().heartbeat_interval));
     }
   });
   heart_beat_thread_->detach();
@@ -326,7 +326,7 @@ void AbstractNode::UpdateSchedulerTime() {
 bool AbstractNode::CheckSchedulerTimeout() const {
   struct timeval current_time {};
   (void)gettimeofday(&current_time, nullptr);
-  if (scheduler_time_.tv_sec + ClusterMetadata::instance()->scheduler_timeout() < current_time.tv_sec) {
+  if (scheduler_time_.tv_sec + PSContext::instance()->cluster_config().scheduler_timeout < current_time.tv_sec) {
     return true;
   }
   return false;
@@ -411,8 +411,8 @@ bool AbstractNode::WaitForDisconnect(const uint32_t &timeout) {
 }
 
 bool AbstractNode::InitClientToScheduler() {
-  std::string scheduler_host = ClusterMetadata::instance()->scheduler_host();
-  uint16_t scheduler_port = ClusterMetadata::instance()->scheduler_port();
+  std::string scheduler_host = PSContext::instance()->cluster_config().scheduler_host;
+  uint16_t scheduler_port = PSContext::instance()->cluster_config().scheduler_port;
   client_to_scheduler_ = std::make_shared<TcpClient>(scheduler_host, scheduler_port);
   client_to_scheduler_->SetMessageCallback(
     [&](std::shared_ptr<MessageMeta> meta, const Protos &protos, const void *data, size_t size) {
@@ -438,7 +438,7 @@ bool AbstractNode::InitClientToScheduler() {
   client_to_scheduler_thread_->detach();
 
   client_to_scheduler_->set_disconnected_callback([&]() {
-    std::this_thread::sleep_for(std::chrono::milliseconds(ClusterMetadata::instance()->connect_interval()));
+    std::this_thread::sleep_for(std::chrono::milliseconds(PSContext::instance()->cluster_config().connect_interval));
     if (is_ready_.load() == false) {
       client_to_scheduler_->Init();
     }

mindspore/ccsrc/ps/core/cluster_config.h

@@ -0,0 +1,73 @@
+/**
+ * Copyright 2021 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef MINDSPORE_CCSRC_PS_CORE_CLUSTER_CONFIG_H_
+#define MINDSPORE_CCSRC_PS_CORE_CLUSTER_CONFIG_H_
+
+#include <string>
+#include <iostream>
+#include <memory>
+#include <utility>
+
+#include "utils/log_adapter.h"
+
+namespace mindspore {
+namespace ps {
+namespace core {
+/*
+ * Configuration information read through environment variables and configuration files, generally immutable
+ */
+struct ClusterConfig {
+  ClusterConfig()
+      : initial_worker_num(0),
+        initial_server_num(0),
+        heartbeat_interval(3),
+        scheduler_host(""),
+        scheduler_port(0),
+        heartbeat_timeout(30),
+        cluster_available_timeout(300),
+        connect_interval(100),
+        scheduler_timeout(30) {}
+
+  void Init(const uint32_t &worker_num, const uint32_t &server_num, std::string host, const uint16_t &port) {
+    initial_worker_num = worker_num;
+    initial_server_num = server_num;
+    scheduler_host = host;
+    scheduler_port = port;
+  }
+
+  // Configure through environment variables:MS_WORKER_NUM
+  uint32_t initial_worker_num;
+  // Configure through environment variables:MS_SERVER_NUM
+  uint32_t initial_server_num;
+
+  // The interval for sending heartbeat packets between worker node,server node and scheduler node is 3 seconds.
+  uint32_t heartbeat_interval;
+  std::string scheduler_host;
+  uint16_t scheduler_port;
+  // The timeout for worker node and server node sending heartbeat packets to scheduler node is 30 seconds.
+  uint32_t heartbeat_timeout;
+  // Timeout period for cluster preparation is 300 seconds.
+  uint32_t cluster_available_timeout;
+  // The timeout period for the client to connect to the server is 100ms.
+  uint32_t connect_interval;
+  // When the scheduler exits, the worker and server can continue to work for 5 hours
+  uint32_t scheduler_timeout;
+};
+}  // namespace core
+}  // namespace ps
+}  // namespace mindspore
+#endif  // MINDSPORE_CCSRC_PS_CORE_CLUSTER_CONFIG_H_

mindspore/ccsrc/ps/core/cluster_metadata.cc

@@ -1,79 +0,0 @@
-/**
- * Copyright 2021 Huawei Technologies Co., Ltd
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "ps/core/cluster_metadata.h"
-#include <string>
-
-namespace mindspore {
-namespace ps {
-namespace core {
-std::shared_ptr<ClusterMetadata> ClusterMetadata::instance() {
-  static std::shared_ptr<ClusterMetadata> metadata_instance = nullptr;
-  if (metadata_instance == nullptr) {
-    metadata_instance.reset(new (std::nothrow) ClusterMetadata());
-  }
-  return metadata_instance;
-}
-
-void ClusterMetadata::Init(const uint32_t &worker_num, const uint32_t &server_num, std::string scheduler_host,
-                           const uint16_t &scheduler_port) {
-  worker_num_ = worker_num;
-  server_num_ = server_num;
-  scheduler_host_ = std::make_unique<std::string>(scheduler_host);
-  scheduler_port_ = scheduler_port;
-}
-
-uint32_t ClusterMetadata::total_worker_num() { return worker_num_; }
-
-uint32_t ClusterMetadata::total_server_num() { return server_num_; }
-
-uint32_t ClusterMetadata::heartbeat_interval() { return heartbeat_interval_; }
-
-void ClusterMetadata::set_heartbeat_interval(const uint32_t &heartbeat_interval) {
-  heartbeat_interval_ = heartbeat_interval;
-}
-
-std::string ClusterMetadata::scheduler_host() {
-  MS_EXCEPTION_IF_NULL(scheduler_host_);
-  return *scheduler_host_;
-}
-
-uint16_t ClusterMetadata::scheduler_port() { return scheduler_port_; }
-
-uint32_t ClusterMetadata::heartbeat_timeout() { return heartbeat_timeout_; }
-
-void ClusterMetadata::set_heartbeat_timeout(const uint32_t &heartbeat_timeout) {
-  heartbeat_interval_ = heartbeat_timeout;
-}
-
-uint32_t ClusterMetadata::cluster_available_timeout() { return cluster_available_timeout_; }
-
-void ClusterMetadata::set_cluster_available_timeout(const uint32_t &cluster_available_timeout) {
-  cluster_available_timeout_ = cluster_available_timeout;
-}
-
-uint32_t ClusterMetadata::connect_interval() { return connect_interval_; }
-
-void ClusterMetadata::set_connect_interval(const uint32_t &connect_interval) { connect_interval_ = connect_interval; }
-
-uint32_t ClusterMetadata::scheduler_timeout() { return scheduler_timeout_; }
-
-void ClusterMetadata::set_scheduler_timeout(const uint32_t &scheduler_timeout) {
-  scheduler_timeout_ = scheduler_timeout;
-}
-}  // namespace core
-}  // namespace ps
-}  // namespace mindspore

mindspore/ccsrc/ps/core/cluster_metadata.h

@@ -27,55 +27,19 @@
 namespace mindspore {
 namespace ps {
 namespace core {
-class ClusterMetadata {
- public:
-  ~ClusterMetadata() = default;
-  ClusterMetadata(ClusterMetadata const &) = delete;
-  ClusterMetadata &operator=(const ClusterMetadata &) = delete;
-  static std::shared_ptr<ClusterMetadata> instance();
+/*
+ * The metadata information of the cluster, stored in the scheduler, is generally used for scale out and scale in.
+ */
+struct ClusterMetadata {
+  ClusterMetadata() : worker_num_(0), server_num_(0) {}
 
-  void Init(const uint32_t &worker_num, const uint32_t &server_num, std::string scheduler_host,
-            const uint16_t &scheduler_port);
-  uint32_t total_worker_num();
-  uint32_t total_server_num();
-  uint32_t heartbeat_interval();
-  void set_heartbeat_interval(const uint32_t &heartbeat_interval);
-  std::string scheduler_host();
-  uint16_t scheduler_port();
-  uint32_t heartbeat_timeout();
-  void set_heartbeat_timeout(const uint32_t &heartbeat_timeout);
-  uint32_t cluster_available_timeout();
-  void set_cluster_available_timeout(const uint32_t &cluster_available_timeout);
-  uint32_t connect_interval();
-  void set_connect_interval(const uint32_t &connect_interval);
-  uint32_t scheduler_timeout();
-  void set_scheduler_timeout(const uint32_t &scheduler_timeout);
+  void Init(const uint32_t &worker_num, const uint32_t &server_num) {
+    worker_num_ = worker_num;
+    server_num_ = server_num;
+  }
 
- private:
-  ClusterMetadata()
-      : worker_num_(0),
-        server_num_(0),
-        heartbeat_interval_(3),
-        scheduler_host_(nullptr),
-        scheduler_port_(0),
-        heartbeat_timeout_(30),
-        cluster_available_timeout_(300),
-        connect_interval_(100),
-        scheduler_timeout_(30) {}
   uint32_t worker_num_;
   uint32_t server_num_;
-  // The interval for sending heartbeat packets between worker node,server node and scheduler node is 3 seconds.
-  uint32_t heartbeat_interval_;
-  std::unique_ptr<std::string> scheduler_host_;
-  uint16_t scheduler_port_;
-  // The timeout for worker node and server node sending heartbeat packets to scheduler node is 30 seconds.
-  uint32_t heartbeat_timeout_;
-  // Timeout period for cluster preparation is 300 seconds.
-  uint32_t cluster_available_timeout_;
-  // The timeout period for the client to connect to the server is 100ms.
-  uint32_t connect_interval_;
-  // When the scheduler exits, the worker and server can continue to work for 5 hours
-  uint32_t scheduler_timeout_;
 };
 }  // namespace core
 }  // namespace ps

mindspore/ccsrc/ps/core/comm_util.cc

@@ -122,9 +122,10 @@ std::string CommUtil::NodeRoleToString(const NodeRole &role) {
   }
 }
 bool CommUtil::ValidateRankId(const enum NodeRole &node_role, const uint32_t &rank_id) {
-  if (node_role == NodeRole::SERVER && (rank_id > ClusterMetadata::instance()->total_server_num() - 1)) {
+  if (node_role == NodeRole::SERVER && (rank_id > PSContext::instance()->cluster_config().initial_server_num - 1)) {
     return false;
-  } else if (node_role == NodeRole::WORKER && (rank_id > ClusterMetadata::instance()->total_worker_num() - 1)) {
+  } else if (node_role == NodeRole::WORKER &&
+             (rank_id > PSContext::instance()->cluster_config().initial_worker_num - 1)) {
     return false;
   }
   return true;

mindspore/ccsrc/ps/core/comm_util.h

@@ -50,6 +50,8 @@
 #include "proto/comm.pb.h"
 #include "proto/ps.pb.h"
 #include "ps/core/cluster_metadata.h"
+#include "ps/core/cluster_config.h"
+#include "ps/ps_context.h"
 #include "utils/log_adapter.h"
 
 namespace mindspore {

mindspore/ccsrc/ps/core/communicator/tcp_client.h

@@ -34,6 +34,7 @@
 #include <condition_variable>
 
 #include "ps/core/cluster_metadata.h"
+#include "ps/core/cluster_config.h"
 #include "utils/convert_utils_base.h"
 #include "ps/core/comm_util.h"
 #include "ps/core/communicator/ssl_wrapper.h"
@@ -58,7 +59,8 @@ class TcpClient {
   std::string GetServerAddress() const;
   void set_disconnected_callback(const OnDisconnected &disconnected);
   void set_connected_callback(const OnConnected &connected);
-  bool WaitConnected(const uint32_t &connected_timeout = ClusterMetadata::instance()->cluster_available_timeout());
+  bool WaitConnected(
+    const uint32_t &connected_timeout = PSContext::instance()->cluster_config().cluster_available_timeout);
   void Init();
   void StartWithDelay(int seconds);
   void Stop();

mindspore/ccsrc/ps/core/communicator/tcp_communicator.h

@@ -25,6 +25,7 @@
 #include "proto/ps.pb.h"
 #include "ps/core/server_node.h"
 #include "ps/core/cluster_metadata.h"
+#include "ps/core/cluster_config.h"
 #include "ps/ps_context.h"
 #include "ps/core/communicator/task_executor.h"
 #include "ps/core/communicator/communicator_base.h"

mindspore/ccsrc/ps/core/communicator/tcp_server.h

@@ -38,6 +38,7 @@
 #include "ps/core/communicator/tcp_message_handler.h"
 #include "ps/core/communicator/ssl_wrapper.h"
 #include "ps/core/cluster_metadata.h"
+#include "ps/core/cluster_config.h"
 #include "utils/convert_utils_base.h"
 #include "ps/core/comm_util.h"
 #include "ps/constants.h"

mindspore/ccsrc/ps/core/node.h

@@ -31,6 +31,8 @@
 #include <tuple>
 
 #include "ps/core/cluster_metadata.h"
+#include "ps/core/cluster_config.h"
+#include "ps/ps_context.h"
 #include "ps/core/node_info.h"
 #include "ps/core/communicator/tcp_client.h"
 #include "ps/core/communicator/tcp_server.h"
@@ -54,7 +56,7 @@ class Node {
   using OnNodeEventMessage = std::function<void(const NodeEvent &event)>;
   using MessageCallback = std::function<void()>;
 
-  virtual bool Start(const uint32_t &timeout = ClusterMetadata::instance()->cluster_available_timeout()) = 0;
+  virtual bool Start(const uint32_t &timeout = PSContext::instance()->cluster_config().cluster_available_timeout) = 0;
   virtual bool Stop() = 0;
   virtual bool Finish(const uint32_t &timeout = kTimeoutInSeconds) = 0;
 

mindspore/ccsrc/ps/core/node_manager.cc

@@ -20,7 +20,8 @@ namespace mindspore {
 namespace ps {
 namespace core {
 void NodeManager::InitNodeNum() {
-  total_node_num_ = ClusterMetadata::instance()->total_server_num() + ClusterMetadata::instance()->total_worker_num();
+  total_node_num_ = PSContext::instance()->cluster_config().initial_server_num +
+                    PSContext::instance()->cluster_config().initial_worker_num;
 }
 
 int NodeManager::NextRankId(const RegisterMessage &register_message) {
@@ -39,7 +40,7 @@ int NodeManager::NextRankId(const RegisterMessage &register_message) {
     uint32_t port = register_message.port();
 
     rank_id = ++next_server_rank_id_;
-    if (IntToUint(rank_id) >= ClusterMetadata::instance()->total_server_num()) {
+    if (IntToUint(rank_id) >= PSContext::instance()->cluster_config().initial_server_num) {
       MS_LOG(WARNING) << "The rank id is greater than the number of servers.";
       rank_id = -1;
       --next_server_rank_id_;
@@ -55,7 +56,7 @@ int NodeManager::NextRankId(const RegisterMessage &register_message) {
                  << " assign rank id:" << rank_id;
   } else if (register_message.role() == NodeRole::WORKER) {
     rank_id = ++next_worker_rank_id_;
-    if (IntToUint(rank_id) >= ClusterMetadata::instance()->total_worker_num()) {
+    if (IntToUint(rank_id) >= PSContext::instance()->cluster_config().initial_worker_num) {
       MS_LOG(WARNING) << "The rank id is greater than the number of workers.";
       rank_id = -1;
       --next_worker_rank_id_;
@@ -104,7 +105,7 @@ void NodeManager::UpdateClusterState() {
   (void)gettimeofday(&current_time, nullptr);
   timeout_nodes_info_.clear();
   for (auto it = heartbeats_.begin(); it != heartbeats_.end(); ++it) {
-    if (it->second.tv_sec + ClusterMetadata::instance()->heartbeat_timeout() < current_time.tv_sec) {
+    if (it->second.tv_sec + PSContext::instance()->cluster_config().heartbeat_timeout < current_time.tv_sec) {
       MS_LOG(WARNING) << "The node id:" << it->first << " is timeout!";
       timeout_nodes_info_[it->first] = nodes_info_[it->first];
     }
@@ -130,7 +131,8 @@ void NodeManager::UpdateClusterState() {
 
 void NodeManager::CheckClusterTimeout() {
   if (total_node_num_ != nodes_info_.size()) {
-    MS_LOG(WARNING) << "The cluster is not ready after " << ClusterMetadata::instance()->cluster_available_timeout()
+    MS_LOG(WARNING) << "The cluster is not ready after "
+                    << PSContext::instance()->cluster_config().cluster_available_timeout
                     << " seconds,so finish the cluster, and change total node number from " << total_node_num_ << " to "
                     << nodes_info_.size();
     current_node_num_ = nodes_info_.size();

mindspore/ccsrc/ps/core/scheduler_node.cc

@@ -87,8 +87,8 @@ void SchedulerNode::InitCommandHandler() {
 void SchedulerNode::CreateTcpServer() {
   node_manager_.InitNodeNum();
 
-  std::string scheduler_host = ClusterMetadata::instance()->scheduler_host();
-  uint32_t scheduler_port = ClusterMetadata::instance()->scheduler_port();
+  std::string scheduler_host = PSContext::instance()->cluster_config().scheduler_host;
+  uint32_t scheduler_port = PSContext::instance()->cluster_config().scheduler_port;
   server_ = std::make_shared<TcpServer>(scheduler_host, scheduler_port);
   server_->SetMessageCallback([&](std::shared_ptr<TcpConnection> conn, std::shared_ptr<MessageMeta> meta,
                                   const Protos &protos, const void *data, size_t size) {
@@ -168,19 +168,20 @@ void SchedulerNode::StartUpdateClusterStateTimer() {
       // 1. update cluster timeout
       if (!node_manager_.is_cluster_ready() &&
           (std::chrono::steady_clock::now() - start_time >
-           std::chrono::seconds(ClusterMetadata::instance()->cluster_available_timeout()))) {
+           std::chrono::seconds(PSContext::instance()->cluster_config().cluster_available_timeout))) {
         node_manager_.CheckClusterTimeout();
       }
 
       // 2. update cluster state
-      std::this_thread::sleep_for(std::chrono::seconds(ClusterMetadata::instance()->heartbeat_interval()));
+      std::this_thread::sleep_for(std::chrono::seconds(PSContext::instance()->cluster_config().heartbeat_interval));
       node_manager_.UpdateClusterState();
       if (node_manager_.is_cluster_ready()) {
         is_ready_ = true;
         wait_start_cond_.notify_all();
       }
       if (node_manager_.is_cluster_finish()) {
-        std::this_thread::sleep_for(std::chrono::seconds(ClusterMetadata::instance()->heartbeat_interval() * 2));
+        std::this_thread::sleep_for(
+          std::chrono::seconds(PSContext::instance()->cluster_config().heartbeat_interval * 2));
         is_finish_ = true;
         wait_finish_cond_.notify_all();
       }

mindspore/ccsrc/ps/core/scheduler_node.h

@@ -28,6 +28,8 @@
 #include <unordered_map>
 
 #include "ps/core/cluster_metadata.h"
+#include "ps/core/cluster_config.h"
+#include "ps/ps_context.h"
 #include "ps/core/communicator/tcp_client.h"
 #include "ps/core/communicator/tcp_server.h"
 #include "ps/core/node_manager.h"
@@ -44,7 +46,7 @@ class SchedulerNode : public Node {
   typedef void (SchedulerNode::*ResponseHandler)(std::shared_ptr<TcpServer> server, std::shared_ptr<TcpConnection> conn,
                                                  std::shared_ptr<MessageMeta> meta, const void *data, size_t size);
 
-  bool Start(const uint32_t &timeout = ClusterMetadata::instance()->cluster_available_timeout()) override;
+  bool Start(const uint32_t &timeout = PSContext::instance()->cluster_config().cluster_available_timeout) override;
   bool Stop() override;
   bool Finish(const uint32_t &timeout = kTimeoutInSeconds) override;
 

mindspore/ccsrc/ps/core/server_node.cc

@@ -147,7 +147,6 @@ std::shared_ptr<CommunicatorBase> ServerNode::GetOrCreateTcpComm(const std::stri
     MS_LOG(INFO) << "Create Tcp communicator.";
     auto tcp_comm = std::make_shared<TcpCommunicator>(task_executor, this);
     MS_EXCEPTION_IF_NULL(tcp_comm);
-    ClusterMetadata::instance()->Init(worker_num, server_num, scheduler_ip, scheduler_port);
     MS_LOG(INFO) << "Initialize cluster metadata for server. Worker number:" << worker_num
                  << ", Server number:" << server_num << ", Scheduler ip:" << scheduler_ip
                  << ", Scheduler port:" << scheduler_port;

mindspore/ccsrc/ps/core/server_node.h

@@ -27,6 +27,8 @@
 #include <unordered_map>
 
 #include "ps/core/cluster_metadata.h"
+#include "ps/core/cluster_config.h"
+#include "ps/ps_context.h"
 #include "ps/core/communicator/tcp_client.h"
 #include "ps/core/communicator/tcp_server.h"
 #include "ps/core/abstract_node.h"
@@ -45,7 +47,7 @@ class ServerNode : public AbstractNode {
 
   ~ServerNode() override = default;
 
-  bool Start(const uint32_t &timeout = ClusterMetadata::instance()->cluster_available_timeout()) override;
+  bool Start(const uint32_t &timeout = PSContext::instance()->cluster_config().cluster_available_timeout) override;
   bool Stop() override;
   bool Finish(const uint32_t &timeout = kTimeoutInSeconds) override;
 

mindspore/ccsrc/ps/core/worker_node.h

@@ -25,6 +25,8 @@
 #include <algorithm>
 
 #include "ps/core/cluster_metadata.h"
+#include "ps/core/cluster_config.h"
+#include "ps/ps_context.h"
 #include "ps/core/communicator/tcp_client.h"
 #include "ps/core/communicator/tcp_server.h"
 #include "ps/core/abstract_node.h"
@@ -37,7 +39,7 @@ class WorkerNode : public AbstractNode {
   WorkerNode() = default;
   ~WorkerNode() override = default;
 
-  bool Start(const uint32_t &timeout = ClusterMetadata::instance()->cluster_available_timeout()) override;
+  bool Start(const uint32_t &timeout = PSContext::instance()->cluster_config().cluster_available_timeout) override;
   bool Stop() override;
   bool Finish(const uint32_t &timeout = kTimeoutInSeconds) override;
 

mindspore/ccsrc/ps/ps_context.cc

@@ -52,7 +52,7 @@ void PSContext::SetPSEnable(bool enabled) {
     server_num_ = std::strtol(common::GetEnv(kEnvPServerNum).c_str(), nullptr, 10);
     scheduler_host_ = common::GetEnv(kEnvSchedulerHost);
     scheduler_port_ = std::strtol(common::GetEnv(kEnvSchedulerPort).c_str(), nullptr, 10);
-    core::ClusterMetadata::instance()->Init(worker_num_, server_num_, scheduler_host_, scheduler_port_);
+    cluster_config_.Init(worker_num_, server_num_, scheduler_host_, scheduler_port_);
   } else {
     MS_LOG(INFO) << "PS mode is disabled.";
     is_worker_ = false;
@@ -311,5 +311,7 @@ bool PSContext::secure_aggregation() const { return secure_aggregation_; }
 bool PSContext::enable_ssl() const { return enable_ssl_; }
 
 void PSContext::set_enable_ssl(bool enabled) { enable_ssl_ = enabled; }
+
+core::ClusterConfig &PSContext::cluster_config() { return cluster_config_; }
 }  // namespace ps
 }  // namespace mindspore

mindspore/ccsrc/ps/ps_context.h

@@ -22,6 +22,7 @@
 #include <memory>
 #include "ps/constants.h"
 #include "ps/core/cluster_metadata.h"
+#include "ps/core/cluster_config.h"
 
 namespace mindspore {
 namespace ps {
@@ -147,6 +148,8 @@ class PSContext {
   void set_secure_aggregation(bool secure_aggregation);
   bool secure_aggregation() const;
 
+  core::ClusterConfig &cluster_config();
+
  private:
   PSContext()
       : ps_enabled_(false),
@@ -229,6 +232,9 @@ class PSContext {
 
   // Whether to use secure aggregation algorithm. Used in federated learning for now.
   bool secure_aggregation_;
+
+  // The cluster config read through environment variables, the value does not change.
+  core::ClusterConfig cluster_config_;
 };
 }  // namespace ps
 }  // namespace mindspore

mindspore/ccsrc/ps/scheduler.cc

@@ -20,9 +20,6 @@ namespace mindspore {
 namespace ps {
 void Scheduler::Run() {
   MS_LOG(INFO) << "Start scheduler.";
-  core::ClusterMetadata::instance()->Init(
-    PSContext::instance()->initial_worker_num(), PSContext::instance()->initial_server_num(),
-    PSContext::instance()->scheduler_host(), PSContext::instance()->scheduler_port());
   scheduler_node_.Start();
   scheduler_node_.Finish();
   scheduler_node_.Stop();

tests/ut/cpp/ps/core/cluster_available_timeout_test.cc

@@ -31,12 +31,9 @@ class TestClusterAvailableTimeout : public UT::Common {
 };
 
 TEST_F(TestClusterAvailableTimeout, TestClusterAvailableTimeout) {
-  ClusterMetadata::instance()->Init(1, 1, "127.0.0.1", 9999);
-  ClusterMetadata::instance()->set_cluster_available_timeout(3);
+  PSContext::instance()->cluster_config().Init(1, 1, "127.0.0.1", 9999);
+  MS_LOG(INFO) << "The timeout is:" << PSContext::instance()->cluster_config().cluster_available_timeout;
   SchedulerNode node;
-  node.Start();
-  node.Finish();
-  node.Stop();
 }
 }  // namespace core
 }  // namespace ps

tests/ut/cpp/ps/core/cluster_metadata_test.cc

@@ -19,26 +19,24 @@
 
 #include "common/common_test.h"
 #include "ps/core/cluster_metadata.h"
+#include "ps/core/cluster_config.h"
+#include "ps/ps_context.h"
 
 namespace mindspore {
 namespace ps {
 namespace core {
-class TestClusterMetadata : public UT::Common {
+class TestClusterConfig : public UT::Common {
  public:
-  TestClusterMetadata() = default;
-  virtual ~TestClusterMetadata() = default;
+  TestClusterConfig() = default;
+  virtual ~TestClusterConfig() = default;
 
   void SetUp() override {}
   void TearDown() override {}
 };
 
-TEST_F(TestClusterMetadata, HeartbeatInterval) {
-  ClusterMetadata::instance()->Init(2, 2, "127.0.0.1", 8080);
-  EXPECT_TRUE(ClusterMetadata::instance()->heartbeat_interval() == 3);
-  ClusterMetadata::instance()->set_heartbeat_interval(100);
-  EXPECT_TRUE(ClusterMetadata::instance()->heartbeat_interval() == 100);
-  EXPECT_STREQ(ClusterMetadata::instance()->scheduler_host().c_str(), "127.0.0.1");
-  EXPECT_TRUE(ClusterMetadata::instance()->scheduler_port() == 8080);
+TEST_F(TestClusterConfig, HeartbeatInterval) {
+  PSContext::instance()->cluster_config().Init(2, 2, "127.0.0.1", 8080);
+  PSContext::instance()->cluster_config().heartbeat_interval = 100;
 }
 }  // namespace core
 }  // namespace ps

tests/ut/cpp/ps/core/common_util_test.cc

@@ -53,7 +53,7 @@ TEST_F(TestCommUtil, GetAvailableInterfaceAndIP) {
 }
 
 TEST_F(TestCommUtil, ValidateRankId) {
-  ClusterMetadata::instance()->Init(3, 2, "127.0.0.1", 9999);
+  PSContext::instance()->cluster_config().Init(3, 2, "127.0.0.1", 9999);
   EXPECT_TRUE(CommUtil::ValidateRankId(NodeRole::WORKER, 2));
   EXPECT_FALSE(CommUtil::ValidateRankId(NodeRole::WORKER, 3));
   EXPECT_TRUE(CommUtil::ValidateRankId(NodeRole::SERVER, 1));