dtrace discuss - Oct 2009 - Java hotspot monitor-contended-enter probe

By reading manual, I have an impression that this probe is fired when a thread
enters a synchronized block, but in my simple example it did not.

<p>If this is not the right probe for it which one I have to use?

<p>Thanks,
   Tigran.


<p>P.S: my test app
<pre>
import java.util.concurrent.TimeUnit;
import java.util.Vector;
public class a {

   public static void main(String[] args) throws Exception  {
      Vector v = new Vector();
      final Object lock = new Object();
      for(int i = 0; i &lt; 100000; i++) {
          synchronized(lock) {
             v.add(i);
          }
             try {
                 TimeUnit.SECONDS.sleep(2);
                 System.out.print(".");
             } catch (InterruptedException ex) {
                 ex.printStackTrace();
             }

      }
      System.out.println("\n");
   }
}
</pre>
-- 
This message posted from opensolaris.org

This probe is fired only when the thread tries to enter a contended
monitor. In your example below, there''s only one thread trying to
acquire the lock, so no contention occur.

Fred


Tigran Mkrtchyan wrote:
> By reading manual, I have an impression that this probe is fired when a
thread enters a synchronized block, but in my simple example it did not.
> 
> <p>If this is not the right probe for it which one I have to use?
> 
> <p>Thanks,
>    Tigran.
> 
> 
> <p>P.S: my test app
> <pre>
> import java.util.concurrent.TimeUnit;
> import java.util.Vector;
> public class a {
> 
>    public static void main(String[] args) throws Exception  {
>       Vector v = new Vector();
>       final Object lock = new Object();
>       for(int i = 0; i &lt; 100000; i++) {
>           synchronized(lock) {
>              v.add(i);
>           }
>              try {
>                  TimeUnit.SECONDS.sleep(2);
>                  System.out.print(".");
>              } catch (InterruptedException ex) {
>                  ex.printStackTrace();
>              }
> 
>       }
>       System.out.println("\n");
>    }
> }
> </pre>
-- 
Frederic Parain - Sun Microsystems
180, av. de l''Europe, ZIRST de Montbonnot
38334 Saint Ismier, FRANCE
Phone: +33 4 76 18 81 17
Email: Frederic.Parain at Sun.COM

In other word only when other thread in synchronized block.
Fair enough, but is there another probe which fired when a thread enters a
synchronized block?

Thanks,
    Tigran.
-- 
This message posted from opensolaris.org

There''s no such probe.
In fact, entering a not contended monitor is very common
is Java, so in this case, the VM uses a highly optimized
fast path which is not instrumented with a DTrace probe.

Fred

Tigran Mkrtchyan wrote:
> In other word only when other thread in synchronized block.
> Fair enough, but is there another probe which fired when a thread enters a
> synchronized block?
> 
> Thanks,
>     Tigran.
-- 
Frederic Parain - Sun Microsystems
180, av. de l''Europe, ZIRST de Montbonnot
38334 Saint Ismier, FRANCE
Phone: +33 4 76 18 81 17
Email: Frederic.Parain at Sun.COM

That''s unfortunate. I was hoping to  write a script for deadlock
detection
by following lock which each thread holds.

Ok, will try to some other ways to do so (may be jvm tools provides some
information).

Thanks a lot,
	Tigran.
-- 
This message posted from opensolaris.org

You should look at jstack.
http://java.sun.com/javase/6/webnotes/trouble/TSG-VM/html/tooldescr.html#gblfh

Fred

Tigran Mkrtchyan wrote:
> That''s unfortunate. I was hoping to  write a script for deadlock
detection
> by following lock which each thread holds.
> 
> Ok, will try to some other ways to do so (may be jvm tools provides some
information).
> 
> Thanks a lot,
> 	Tigran.
-- 
Frederic Parain - Sun Microsystems
180, av. de l''Europe, ZIRST de Montbonnot
38334 Saint Ismier, FRANCE
Phone: +33 4 76 18 81 17
Email: Frederic.Parain at Sun.COM

Tigran Mkrtchyan wrote:
> That''s unfortunate. I was hoping to  write a script for deadlock
detection
> by following lock which each thread holds.
> 
> Ok, will try to some other ways to do so (may be jvm tools provides some
information).
> 
> Thanks a lot,
> 	Tigran.
The JVM already does some deadlock detection, but I''m not sure exactly 
what you''re trying to do so perhaps it''s not usable.

If you looking for actual deadlocks (as opposed to "potential" 
deadlocks), then the contended lock cases are in fact what you want, 
since each lock involved in the deadlock would be contended.

--
- Keith

thanks.
-- 
This message posted from opensolaris.org