Monitoring CPU, RAM, I/O usage at time of Java Garbage Collection

问题

I am using -Xloggc to output the GC messages to a file. However I am also interested in knowing the system parameters like CPU, Memory, I/O when the GC events happen. I understand that sar linux command is there, but how can I know the metrics at time of GC events instead of manually comparing the results using time stamps.

Java 1.7
Oracle enterprise linux 2.6.39

Thanks.

回答1:

If you're up to Java 1.7 update 4, you can register a GarbageCollectorMXBean.

Here's an example of its setup, including emissions concerning memory (you can get more system parameters with MX classes, too):

public static void installGCMonitoring(){
    //get all the GarbageCollectorMXBeans - there's one for each heap generation
    //so probably two - the old generation and young generation
    List<GarbageCollectorMXBean> gcbeans = java.lang.management.ManagementFactory.getGarbageCollectorMXBeans();
    //Install a notifcation handler for each bean
    for (GarbageCollectorMXBean gcbean : gcbeans) {
      System.out.println(gcbean);
      NotificationEmitter emitter = (NotificationEmitter) gcbean;
      //use an anonymously generated listener for this example
      // - proper code should really use a named class
      NotificationListener listener = new NotificationListener() {
        //keep a count of the total time spent in GCs
        long totalGcDuration = 0;

        //implement the notifier callback handler
        @Override
        public void handleNotification(Notification notification, Object handback) {
          //we only handle GARBAGE_COLLECTION_NOTIFICATION notifications here
          if (notification.getType().equals(GarbageCollectionNotificationInfo.GARBAGE_COLLECTION_NOTIFICATION)) {
            //get the information associated with this notification
            GarbageCollectionNotificationInfo info = GarbageCollectionNotificationInfo.from((CompositeData) notification.getUserData());
            //get all the info and pretty print it
            long duration = info.getGcInfo().getDuration();
            String gctype = info.getGcAction();
            if ("end of minor GC".equals(gctype)) {
              gctype = "Young Gen GC";
            } else if ("end of major GC".equals(gctype)) {
              gctype = "Old Gen GC";
            }
            System.out.println();
            System.out.println(gctype + ": - " + info.getGcInfo().getId()+ " " + info.getGcName() + " (from " + info.getGcCause()+") "+duration + " microseconds; start-end times " + info.getGcInfo().getStartTime()+ "-" + info.getGcInfo().getEndTime());
            //System.out.println("GcInfo CompositeType: " + info.getGcInfo().getCompositeType());
            //System.out.println("GcInfo MemoryUsageAfterGc: " + info.getGcInfo().getMemoryUsageAfterGc());
            //System.out.println("GcInfo MemoryUsageBeforeGc: " + info.getGcInfo().getMemoryUsageBeforeGc());

            //Get the information about each memory space, and pretty print it
            Map<String, MemoryUsage> membefore = info.getGcInfo().getMemoryUsageBeforeGc();
            Map<String, MemoryUsage> mem = info.getGcInfo().getMemoryUsageAfterGc();
            for (Entry<String, MemoryUsage> entry : mem.entrySet()) {
              String name = entry.getKey();
              MemoryUsage memdetail = entry.getValue();
              long memInit = memdetail.getInit();
              long memCommitted = memdetail.getCommitted();
              long memMax = memdetail.getMax();
              long memUsed = memdetail.getUsed();
              MemoryUsage before = membefore.get(name);
              long beforepercent = ((before.getUsed()*1000L)/before.getCommitted());
              long percent = ((memUsed*1000L)/before.getCommitted()); //>100% when it gets expanded

              System.out.print(name + (memCommitted==memMax?"(fully expanded)":"(still expandable)") +"used: "+(beforepercent/10)+"."+(beforepercent%10)+"%->"+(percent/10)+"."+(percent%10)+"%("+((memUsed/1048576)+1)+"MB) / ");
            }
            System.out.println();
            totalGcDuration += info.getGcInfo().getDuration();
            long percent = totalGcDuration*1000L/info.getGcInfo().getEndTime();
            System.out.println("GC cumulated overhead "+(percent/10)+"."+(percent%10)+"%");
          }
        }
      };

      //Add the listener
      emitter.addNotificationListener(listener, null, null);
    }
  }

回答2:

You can try VisualVM with Tracer add-on. A couple of screenshots show how it would look like

来源：https://stackoverflow.com/questions/32509184/monitoring-cpu-ram-i-o-usage-at-time-of-java-garbage-collection