[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openjdk-serviceability-dev
Subject:    Excess threads reporting on deadlock
From:       dmytro_sheyko () hotmail ! com (Dmytro Sheyko)
Date:       2011-01-31 16:57:10
Message-ID: SNT137-w12648C9E97414CA1B22D198AE20 () phx ! gbl
[Download RAW message or body]


Keith,

I would be very obliged to you if reopen it.

Deadlock detection code is used in ThreadMXBean.findDeadlockedThread() and \
ThreadMXBean.findMonitorDeadlockedThread() (as well as when VM prints out thread \
information). These JMX methods return ids of deadlocked threads, that is VM actually \
does some work to provide focused information.

And according to javadoc

...a thread is monitor deadlocked if it is part of a cycle in the relation "is \
waiting for an object monitor owned by"...

So, formally, if thread is not in a cycle it is not considered deadlocked and hence \
should not be return by methods above. I.e. this issue is likely bug than RFE.

Also I would like suggest a fix for it
https://bugs.openjdk.java.net/show_bug.cgi?id=100065

Thanks,
Dmytro

CC: david.holmes at Sun.COM; serviceability-dev at openjdk.java.net
From: keith.mcguigan@oracle.com
To: dmytro_sheyko at hotmail.com
Subject: Re: Excess threads reporting on deadlock
Date: Mon, 31 Jan 2011 11:09:13 -0500


You'll have to excuse me while I struggle to remember the details of this one, but I \
believe that the deadlock detection code in the VM only detects the existence of a \
deadlock when it happens, but it doesn't know which threads are a part of it.   Any \
thread that is currently blocked is a potential contributor to the deadlock, so the \
VM prints out information about all the blocked threads so that the developer can \
determine where the deadlock is and how to fix it.     I could be wrong about this -- \
it's been a while since I looked at it. Again, IMO, this ought to be fine, but I \
understand if you've got lots of blocked threads than you might have a lot of extra \
data to slog through.  I can re-open the bug if you like (perhaps as an RFE?), but \
it's not likely to be a high enough priority to be "fixed" anytime in the near \
future.  Unless someone wants to investigate the code to find and suggest a fix \
                (hint,  hint).
--- Keith
On Jan 31, 2011, at 10:20 AM, Dmytro Sheyko wrote:Hi Keith,

Thank you for information about this bug.
May I convince you of the need for fixing it?

"The more the better" approach does not go well here.
Doesn't it make sense to report all threads (including even not blocked ones)?
Analyzing deadlock it's important to know threads that form cycle. The problem is \
almost always in code that these threads execute. Other threads just hinder analysis. \
But if I need them I can use other API that gives me the list of all threads.

However, existing behavior does not even follow "the more the better" approach \
properly. Sometimes the deadlock is reported as (A, B, C), sometimes as (B, C, D) and \
as (B, ?), but never as (A, B, C, D). The set of threads that are reported is not \
clearly defined.

Thanks,
Dmytro

CC: david.holmes at sun.com; serviceability-dev at openjdk.java.net
From: keith.mcguigan@oracle.com
To: dmytro_sheyko at hotmail.com
Subject: Re: Excess threads reporting on deadlock
Date: Mon, 31 Jan 2011 07:57:00 -0500


Hi Dmytro -
This bug was marked 'closed' because the behavior noted is intentional and we do not \
believe it should be changed.  When an error situation such as a deadlock occurs we \
                want the developer to have as much information as possible to help \
                resolve the issue.
--- Keith
On Jan 31, 2011, at 4:52 AM, Dmytro Sheyko wrote:Corrections: I am about this one:
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6850341

From: dmytro_sheyko@hotmail.com
To: david.holmes at sun.com
Subject: RE: Excess threads reporting on deadlock
Date: Mon, 31 Jan 2011 16:48:55 +0700
CC: serviceability-dev at openjdk.java.net

Hi,

I just noticed that the bug report is closed. Can I know why?

Thanks,
Dmytro

> Date: Fri, 5 Jun 2009 03:41:17 +1000
> From: David.Holmes at Sun.COM
> Subject: Re: Excess threads reporting on deadlock
> To: dmytro_sheyko at hotmail.com
> CC: serviceability-dev at openjdk.java.net
> 
> Hi Dmytro,
> 
> Thanks for submitting these. I just wanted to say that a lack of immediate 
> response on the mailing lists does not mean noone is, or will, look at 
> these, but it may be a while before people have the cycles to do so. 
> Certainly this week is not a good week :)
> 
> David Holmes (from JavaOne)
> 
> Dmytro Sheyko wrote:
> > Hi,
> > 
> > Another one patch about deadlock detection:
> > 
> > Excess threads reporting on deadlock
> > https://bugs.openjdk.java.net/show_bug.cgi?id=100065
> > 
> > --
> > Dmytro Sheyko
> > ------------------------------------------------------------------------
> > Invite your mail contacts to join your friends list with Windows Live 
> > Spaces. It's easy! Try it! 
> > <http://spaces.live.com/spacesapi.aspx?wx_action=create&wx_url=/friends.aspx&mkt=en-us>
> > 


 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.openjdk.java.net/pipermail/serviceability-dev/attachments/20110131/0578ca09/attachment-0001.html \



[prev in list] [next in list] [prev in thread] [next in thread]