Apollo Application and Source Code Analysis: Monitor Monitoring - Software Monitoring - Process Survival Monitoring - process_monitor

Table of contents

process

the code

analyze

Get information about a process that can run

Check module information of HMI

Check the monitored components

Check other components

Judging process status UpdateStatus


process

the code

class ProcessMonitor : public RecurrentRunner {
 public:
  ProcessMonitor();
  void RunOnce(const double current_time) override;

 private:
  static void UpdateStatus(
      const std::vector<std::string>& running_processes,
      const apollo::dreamview::ProcessMonitorConfig& config,
      ComponentStatus* status);
};

void ProcessMonitor::RunOnce(const double current_time) {
  // Get running processes.
  std::vector<std::string> running_processes;
  for (const auto& cmd_file : cyber::common::Glob("/proc/*/cmdline")) {
    // Get process command string.
    std::string cmd_string;
    if (cyber::common::GetContent(cmd_file, &cmd_string) &&
        !cmd_string.empty()) {
      // In /proc/<PID>/cmdline, the parts are separated with \0, which will be
      // converted back to whitespaces here.
      std::replace(cmd_string.begin(), cmd_string.end(), '\0', ' ');
      running_processes.push_back(cmd_string);
    }
  }

  auto manager = MonitorManager::Instance();
  const auto& mode = manager->GetHMIMode();

  // Check HMI modules.
  auto* hmi_modules = manager->GetStatus()->mutable_hmi_modules();
  for (const auto& iter : mode.modules()) {
    const std::string& module_name = iter.first;
    const auto& config = iter.second.process_monitor_config();
    UpdateStatus(running_processes, config, &hmi_modules->at(module_name));
  }

  // Check monitored components.
  auto* components = manager->GetStatus()->mutable_components();
  for (const auto& iter : mode.monitored_components()) {
    const std::string& name = iter.first;
    if (iter.second.has_process() &&
        apollo::common::util::ContainsKey(*components, name)) {
      const auto& config = iter.second.process();
      auto* status = components->at(name).mutable_process_status();
      UpdateStatus(running_processes, config, status);
    }
  }

  // Check other components.
  auto* other_components = manager->GetStatus()->mutable_other_components();
  for (const auto& iter : mode.other_components()) {
    const std::string& name = iter.first;
    const auto& config = iter.second;
    UpdateStatus(running_processes, config, &other_components->at(name));
  }
}

analyze

The core processing logic is still in runOnce

Get information about a process that can run

// Get running processes.
  std::vector<std::string> running_processes;
  for (const auto& cmd_file : cyber::common::Glob("/proc/*/cmdline")) {
    // Get process command string.
    std::string cmd_string;
    if (cyber::common::GetContent(cmd_file, &cmd_string) &&
        !cmd_string.empty()) {
      // In /proc/<PID>/cmdline, the parts are separated with \0, which will be
      // converted back to whitespaces here.
      std::replace(cmd_string.begin(), cmd_string.end(), '\0', ' ');
      running_processes.push_back(cmd_string);
    }
  }

First iterate through all the /proc/*/cmdline files. This file describes the process-related information. The * in the middle is the pid. For the specific content of this file, please refer to: https://blog.csdn.net/whatday/article/ details/108897457

The command line parameters of the process can be obtained from the above file, including the startup path of the process (argv[0])

Check module information of HMI

 // Check HMI modules.
  auto* hmi_modules = manager->GetStatus()->mutable_hmi_modules();
  for (const auto& iter : mode.modules()) {
    const std::string& module_name = iter.first;
    const auto& config = iter.second.process_monitor_config();
    UpdateStatus(running_processes, config, &hmi_modules->at(module_name));
  }
  1. Get all HMI configuration modules;
  2. Check and compare the running process checked in the system with the configured running module, and give the final checking status to &hmi_modules->at(module_name)

Check the monitored components

// Check monitored components.
  auto* components = manager->GetStatus()->mutable_components();
  for (const auto& iter : mode.monitored_components()) {
    const std::string& name = iter.first;
    if (iter.second.has_process() &&
        apollo::common::util::ContainsKey(*components, name)) {
      const auto& config = iter.second.process();
      auto* status = components->at(name).mutable_process_status();
      UpdateStatus(running_processes, config, status);
    }
  }
  1. First get all the components;
  2. Traverse the components to be monitored and obtain the config information of the process;
  3. Give to UpdateStatus for final judgment;

Check other components

  auto* other_components = manager->GetStatus()->mutable_other_components();
  for (const auto& iter : mode.other_components()) {
    const std::string& name = iter.first;
    const auto& config = iter.second;
    UpdateStatus(running_processes, config, &other_components->at(name));
  }
}

Same logic as above

Judging process status UpdateStatus

void ProcessMonitor::UpdateStatus(
    const std::vector<std::string>& running_processes,
    const apollo::dreamview::ProcessMonitorConfig& config,
    ComponentStatus* status) {
  status->clear_status();
  for (const std::string& command : running_processes) {
    bool all_keywords_matched = true;
    for (const std::string& keyword : config.command_keywords()) {
      if (command.find(keyword) == std::string::npos) {
        all_keywords_matched = false;
        break;
      }
    }
    if (all_keywords_matched) {
      // Process command keywords are all matched. The process is running.
      SummaryMonitor::EscalateStatus(ComponentStatus::OK, command, status);
      return;
    }
  }
  SummaryMonitor::EscalateStatus(ComponentStatus::FATAL, "", status);
}

This is the core judgment function of the entire process monitoring.

  1. Traverse the running_processes obtained from /proc/*/cmdline before;
  2. Traverse command_keywords from config, and see if a program name (start position) can be found in running_processes, and the named line parameters can match the item
  3. true if found, false if not found

Guess you like

Origin blog.csdn.net/qq_32378713/article/details/128094813