Use single metric multi-value metric for process performance (#199)

AliceO2Group · May 14, 2020 · 008f13a · 008f13a
1 parent 0795fea
commit 008f13a
Show file tree

Hide file tree

Showing 4 changed files with 27 additions and 43 deletions.
diff --git a/include/Monitoring/ProcessMonitor.h b/include/Monitoring/ProcessMonitor.h
@@ -43,11 +43,8 @@ class ProcessMonitor
   /// Default destructor
   ~ProcessMonitor() = default;
 
-  /// Retrieves memory usage (%)
-  Metric getMemoryUsage();
-
-  /// Retrieves CPU usage (%) and number of context switches during the interval
-  std::vector<Metric> getCpuAndContexts();
+  /// Return performance metrics
+  Metric getPerformanceMetrics();
 
  private:
   /// PIDs that are monitored
@@ -64,6 +61,12 @@ class ProcessMonitor
 
   /// Timestamp when process monitoring was executed last time
   std::chrono::high_resolution_clock::time_point mTimeLastRun;
+
+  /// Retrieves memory usage (%)
+  double getMemoryUsage();
+
+  /// Retrieves CPU usage (%) and number of context switches during the interval
+  Metric getCpuAndContexts();
 };
 
 } // namespace monitoring

diff --git a/src/Monitoring.cxx b/src/Monitoring.cxx
@@ -134,36 +134,13 @@ void Monitoring::pushLoop()
   std::this_thread::sleep_for(std::chrono::milliseconds(100));
   while (mMonitorRunning) {
     if (mProcessMonitoringInterval != 0 && (loopCount % (mProcessMonitoringInterval * 10)) == 0) {
-      transmit(mProcessMonitor->getCpuAndContexts());
-#ifdef O2_MONITORING_OS_LINUX
-      transmit(mProcessMonitor->getMemoryUsage());
-#endif
+      transmit(mProcessMonitor->getPerformanceMetrics());
     }
-
-    /*if (mAutoPushInterval != 0 && (loopCount % (mAutoPushInterval * 10)) == 0) {
-      std::vector<Metric> metrics;
-      for (auto metric : mPushStore) {
-        metric.resetTimestamp();
-        metrics.push_back(metric);
-      }
-      transmit(std::move(metrics));
-    }*/
     std::this_thread::sleep_for(std::chrono::milliseconds(100));
     (loopCount >= 600) ? loopCount = 0 : loopCount++;
   }
 }
 
-/*ComplexMetric& Monitoring::getAutoPushMetric(std::string name, unsigned int interval)
-{
-  if (!mMonitorRunning) {
-    mMonitorRunning = true;
-    mMonitorThread = std::thread(&Monitoring::pushLoop, this);
-    mAutoPushInterval = interval;
-  }
-  mPushStore.emplace_back(std::variant<int, std::string, double, uint64_t>{}, name);
-  return mPushStore.back();
-}*/
-
 void Monitoring::transmit(std::vector<Metric>&& metrics)
 {
   for (auto& backend : mBackends) {

diff --git a/src/ProcessMonitor.cxx b/src/ProcessMonitor.cxx
@@ -28,6 +28,15 @@ ProcessMonitor::ProcessMonitor()
 #endif
 }
 
+Metric ProcessMonitor::getPerformanceMetrics()
+{
+  auto metric = getCpuAndContexts();
+#ifdef O2_MONITORING_OS_LINUX
+  metric.addValue(getMemoryUsage(), "memory_pct");
+#endif
+  return metric;
+}
+
 void ProcessMonitor::setTotalMemory()
 {
   std::ifstream memInfo("/proc/meminfo");
@@ -39,7 +48,7 @@ void ProcessMonitor::setTotalMemory()
   mTotalMemory = std::stoi(tokens[1]);
 }
 
-Metric ProcessMonitor::getMemoryUsage()
+double ProcessMonitor::getMemoryUsage()
 {
   std::ifstream statusStream("/proc/self/status");
   std::string rssString;
@@ -52,30 +61,28 @@ Metric ProcessMonitor::getMemoryUsage()
   std::istringstream iss(rssString);
   std::vector<std::string> tokens{std::istream_iterator<std::string>{iss},
                                   std::istream_iterator<std::string>{}};
-  return Metric{(std::stod(tokens[1]) * 100) / mTotalMemory, "memoryUsagePercentage"};
+  return (std::stod(tokens[1]) * 100) / mTotalMemory;
 }
 
-std::vector<Metric> ProcessMonitor::getCpuAndContexts()
+Metric ProcessMonitor::getCpuAndContexts()
 {
-  std::vector<Metric> metrics;
   struct rusage currentUsage;
   getrusage(RUSAGE_SELF, &currentUsage);
   auto timeNow = std::chrono::high_resolution_clock::now();
   double timePassed = std::chrono::duration_cast<std::chrono::microseconds>(timeNow - mTimeLastRun).count();
   if (timePassed < 950) {
     MonLogger::Get() << "[WARN] Do not invoke Process Monitor more frequent then every 1s" << MonLogger::End();
-    return {};
+    return {"processPerformance"};
   }
   double fractionCpuUsed = (currentUsage.ru_utime.tv_sec * 1000000.0 + currentUsage.ru_utime.tv_usec - (mPreviousGetrUsage.ru_utime.tv_sec * 1000000.0 + mPreviousGetrUsage.ru_utime.tv_usec) + currentUsage.ru_stime.tv_sec * 1000000.0 + currentUsage.ru_stime.tv_usec - (mPreviousGetrUsage.ru_stime.tv_sec * 1000000.0 + mPreviousGetrUsage.ru_stime.tv_usec)) / timePassed;
 
-  metrics.emplace_back(Metric{
-    static_cast<double>(std::round(fractionCpuUsed * 100.0 * 100.0) / 100.0), "cpuUsedPercentage"});
-  metrics.emplace_back(Metric{
-    static_cast<uint64_t>(currentUsage.ru_nivcsw - mPreviousGetrUsage.ru_nivcsw), "involuntaryContextSwitches"});
+  Metric metric{"processPerformance"};
+  metric.addValue(static_cast<double>(std::round(fractionCpuUsed * 100.0 * 100.0) / 100.0), "cpu_user_pct");
+  metric.addValue(static_cast<uint64_t>(currentUsage.ru_nivcsw - mPreviousGetrUsage.ru_nivcsw), "involuntary_context_switches");
 
   mTimeLastRun = timeNow;
   mPreviousGetrUsage = currentUsage;
-  return metrics;
+  return metric;
 }
 
 } // namespace monitoring

diff --git a/test/testProcessMonitor.cxx b/test/testProcessMonitor.cxx
@@ -25,10 +25,7 @@ namespace Test
 BOOST_AUTO_TEST_CASE(createProcessMonitor)
 {
   o2::monitoring::ProcessMonitor processMonitor;
-  processMonitor.getCpuAndContexts();
-#ifdef O2_MONITORING_OS_LINUX
-  processMonitor.getMemoryUsage();
-#endif
+  processMonitor.getPerformanceMetrics();
 }
 
 BOOST_AUTO_TEST_CASE(monitorProcess)