|
| Re: [CDO] Hanging (client) [message #94370 is a reply to message #94273] |
Tue, 28 August 2007 05:45   |
Eclipse User |
|
|
|
Originally posted by: stepper.sympedia.de
Simon McDuff schrieb:
> Eike,
>
> Sometimes the server and the client hang...
>
> If I stop the client and restart and a new one.. it is working correctly.
>
> Each time I try to run it in debug mode... it doesn't hang. I only see the
> behavior in release.
>
That's typical for threading problems (deadlocks).
Sometime there's a chance to hit it in debug mode if all tracing is
disabled (or sent to file instead of the console).
> I wasn't sure at the beginning.. but I have that problem for the last 2
> weeks.
>
> Have you seen this behavior in you environment ?
>
Yes, unfortunately I have this phenomenon aproximately once or twice a
month which is really rare and most of the times without debug mode ;-(
This is one of the most urgent issues to fix (probably in Net4j). Since
it occurs so rarely I planned to investigate it after my work with the
cdo mapper is at least so far that users can try a horizontal mapping
with a table per reference.
I can remember of a single place in Net4j where I had the feeling a race
condition could occur.
When I recall this place I'll tell you about.
I know there are some tools available that try to find synchronization
problems through static code analysis.
Do you have experinece with one of these tools?
> I'm using Java sun 1.5_0_10
>
I'm using a jre1.6.0_02 for my IDE and the code is compiled against a
J2SE-1.5 execution environment which is associated with a jdk1.5.0_12
here at the moment.
If you are ever able to hit the deadlock in debug mode it'd be great if
you send me a stack trace of the main thread (not sure if it's easily
possible to copy *all* stack traces from the debugger).
Cheers
/Eike
|
|
|
| Re: [CDO] Hanging (client) [message #94413 is a reply to message #94370] |
Tue, 28 August 2007 09:53 |
| Eclipse User |
|
|
|
Originally posted by: merks.ca.ibm.com
Eike,
I read somewhere that you can generate a thread dump after a deadlock
occurs under Unix/Linux by running kill -QUIT <pid> and under Windows by
hitting Ctrl + Break.
Eike Stepper wrote:
> Simon McDuff schrieb:
>> Eike,
>>
>> Sometimes the server and the client hang...
>>
>> If I stop the client and restart and a new one.. it is working
>> correctly.
>>
>> Each time I try to run it in debug mode... it doesn't hang. I only
>> see the behavior in release.
>>
> That's typical for threading problems (deadlocks).
> Sometime there's a chance to hit it in debug mode if all tracing is
> disabled (or sent to file instead of the console).
>
>> I wasn't sure at the beginning.. but I have that problem for the last
>> 2 weeks.
>>
>> Have you seen this behavior in you environment ?
>>
> Yes, unfortunately I have this phenomenon aproximately once or twice a
> month which is really rare and most of the times without debug mode ;-(
> This is one of the most urgent issues to fix (probably in Net4j).
> Since it occurs so rarely I planned to investigate it after my work
> with the cdo mapper is at least so far that users can try a horizontal
> mapping with a table per reference.
>
> I can remember of a single place in Net4j where I had the feeling a
> race condition could occur.
> When I recall this place I'll tell you about.
>
> I know there are some tools available that try to find synchronization
> problems through static code analysis.
> Do you have experinece with one of these tools?
>
>> I'm using Java sun 1.5_0_10
>>
> I'm using a jre1.6.0_02 for my IDE and the code is compiled against a
> J2SE-1.5 execution environment which is associated with a jdk1.5.0_12
> here at the moment.
>
> If you are ever able to hit the deadlock in debug mode it'd be great
> if you send me a stack trace of the main thread (not sure if it's
> easily possible to copy *all* stack traces from the debugger).
>
> Cheers
> /Eike
>
|
|
|
| Re: [CDO] Hanging (client) [message #94443 is a reply to message #94413] |
Tue, 28 August 2007 10:36 |
| Eclipse User |
|
|
|
Originally posted by: stepper.sympedia.de
Ed,
That sounds really interesting!
What does "generate" mean? Text in the clipboard? A file?
Or do you mean hitting Ctrl+Break in the concole where the deadlocked
process is running?
This at least doesn't work with the IDE console.
Or should I use debug mode?
No, Then I could simply suspend the locked thread.
Hmmm, it's not clear to me...
/Eike
Ed Merks schrieb:
> Eike,
>
> I read somewhere that you can generate a thread dump after a deadlock
> occurs under Unix/Linux by running kill -QUIT <pid> and under Windows
> by hitting Ctrl + Break.
>
>
> Eike Stepper wrote:
>> Simon McDuff schrieb:
>>> Eike,
>>>
>>> Sometimes the server and the client hang...
>>>
>>> If I stop the client and restart and a new one.. it is working
>>> correctly.
>>>
>>> Each time I try to run it in debug mode... it doesn't hang. I only
>>> see the behavior in release.
>>>
>> That's typical for threading problems (deadlocks).
>> Sometime there's a chance to hit it in debug mode if all tracing is
>> disabled (or sent to file instead of the console).
>>
>>> I wasn't sure at the beginning.. but I have that problem for the
>>> last 2 weeks.
>>>
>>> Have you seen this behavior in you environment ?
>>>
>> Yes, unfortunately I have this phenomenon aproximately once or twice
>> a month which is really rare and most of the times without debug mode
>> ;-(
>> This is one of the most urgent issues to fix (probably in Net4j).
>> Since it occurs so rarely I planned to investigate it after my work
>> with the cdo mapper is at least so far that users can try a
>> horizontal mapping with a table per reference.
>>
>> I can remember of a single place in Net4j where I had the feeling a
>> race condition could occur.
>> When I recall this place I'll tell you about.
>>
>> I know there are some tools available that try to find
>> synchronization problems through static code analysis.
>> Do you have experinece with one of these tools?
>>
>>> I'm using Java sun 1.5_0_10
>>>
>> I'm using a jre1.6.0_02 for my IDE and the code is compiled against a
>> J2SE-1.5 execution environment which is associated with a jdk1.5.0_12
>> here at the moment.
>>
>> If you are ever able to hit the deadlock in debug mode it'd be great
>> if you send me a stack trace of the main thread (not sure if it's
>> easily possible to copy *all* stack traces from the debugger).
>>
>> Cheers
>> /Eike
>>
|
|
|
| Re: [CDO] Hanging (client) [message #94459 is a reply to message #94443] |
Tue, 28 August 2007 10:41 |
| Eclipse User |
|
|
|
Originally posted by: stepper.sympedia.de
http://java.sun.com/developer/technicalArticles/Programming/ Stacktrace/
;-)
Eike Stepper schrieb:
> Ed,
>
> That sounds really interesting!
> What does "generate" mean? Text in the clipboard? A file?
>
> Or do you mean hitting Ctrl+Break in the concole where the deadlocked
> process is running?
> This at least doesn't work with the IDE console.
>
> Or should I use debug mode?
> No, Then I could simply suspend the locked thread.
>
> Hmmm, it's not clear to me...
> /Eike
>
>
> Ed Merks schrieb:
>> Eike,
>>
>> I read somewhere that you can generate a thread dump after a deadlock
>> occurs under Unix/Linux by running kill -QUIT <pid> and under Windows
>> by hitting Ctrl + Break.
>>
>>
>> Eike Stepper wrote:
>>> Simon McDuff schrieb:
>>>> Eike,
>>>>
>>>> Sometimes the server and the client hang...
>>>>
>>>> If I stop the client and restart and a new one.. it is working
>>>> correctly.
>>>>
>>>> Each time I try to run it in debug mode... it doesn't hang. I only
>>>> see the behavior in release.
>>>>
>>> That's typical for threading problems (deadlocks).
>>> Sometime there's a chance to hit it in debug mode if all tracing is
>>> disabled (or sent to file instead of the console).
>>>
>>>> I wasn't sure at the beginning.. but I have that problem for the
>>>> last 2 weeks.
>>>>
>>>> Have you seen this behavior in you environment ?
>>>>
>>> Yes, unfortunately I have this phenomenon aproximately once or twice
>>> a month which is really rare and most of the times without debug
>>> mode ;-(
>>> This is one of the most urgent issues to fix (probably in Net4j).
>>> Since it occurs so rarely I planned to investigate it after my work
>>> with the cdo mapper is at least so far that users can try a
>>> horizontal mapping with a table per reference.
>>>
>>> I can remember of a single place in Net4j where I had the feeling a
>>> race condition could occur.
>>> When I recall this place I'll tell you about.
>>>
>>> I know there are some tools available that try to find
>>> synchronization problems through static code analysis.
>>> Do you have experinece with one of these tools?
>>>
>>>> I'm using Java sun 1.5_0_10
>>>>
>>> I'm using a jre1.6.0_02 for my IDE and the code is compiled against
>>> a J2SE-1.5 execution environment which is associated with a
>>> jdk1.5.0_12 here at the moment.
>>>
>>> If you are ever able to hit the deadlock in debug mode it'd be great
>>> if you send me a stack trace of the main thread (not sure if it's
>>> easily possible to copy *all* stack traces from the debugger).
>>>
>>> Cheers
>>> /Eike
>>>
|
|
|
| Re: [CDO] Hanging (client) [message #94473 is a reply to message #94443] |
Tue, 28 August 2007 10:53 |
| Eclipse User |
|
|
|
Originally posted by: stepper.sympedia.de
This one was also useful:
http://wiki.eclipse.org/How_to_report_a_deadlock
Eike Stepper schrieb:
> Ed,
>
> That sounds really interesting!
> What does "generate" mean? Text in the clipboard? A file?
>
> Or do you mean hitting Ctrl+Break in the concole where the deadlocked
> process is running?
> This at least doesn't work with the IDE console.
>
> Or should I use debug mode?
> No, Then I could simply suspend the locked thread.
>
> Hmmm, it's not clear to me...
> /Eike
>
>
> Ed Merks schrieb:
>> Eike,
>>
>> I read somewhere that you can generate a thread dump after a deadlock
>> occurs under Unix/Linux by running kill -QUIT <pid> and under Windows
>> by hitting Ctrl + Break.
>>
>>
>> Eike Stepper wrote:
>>> Simon McDuff schrieb:
>>>> Eike,
>>>>
>>>> Sometimes the server and the client hang...
>>>>
>>>> If I stop the client and restart and a new one.. it is working
>>>> correctly.
>>>>
>>>> Each time I try to run it in debug mode... it doesn't hang. I only
>>>> see the behavior in release.
>>>>
>>> That's typical for threading problems (deadlocks).
>>> Sometime there's a chance to hit it in debug mode if all tracing is
>>> disabled (or sent to file instead of the console).
>>>
>>>> I wasn't sure at the beginning.. but I have that problem for the
>>>> last 2 weeks.
>>>>
>>>> Have you seen this behavior in you environment ?
>>>>
>>> Yes, unfortunately I have this phenomenon aproximately once or twice
>>> a month which is really rare and most of the times without debug
>>> mode ;-(
>>> This is one of the most urgent issues to fix (probably in Net4j).
>>> Since it occurs so rarely I planned to investigate it after my work
>>> with the cdo mapper is at least so far that users can try a
>>> horizontal mapping with a table per reference.
>>>
>>> I can remember of a single place in Net4j where I had the feeling a
>>> race condition could occur.
>>> When I recall this place I'll tell you about.
>>>
>>> I know there are some tools available that try to find
>>> synchronization problems through static code analysis.
>>> Do you have experinece with one of these tools?
>>>
>>>> I'm using Java sun 1.5_0_10
>>>>
>>> I'm using a jre1.6.0_02 for my IDE and the code is compiled against
>>> a J2SE-1.5 execution environment which is associated with a
>>> jdk1.5.0_12 here at the moment.
>>>
>>> If you are ever able to hit the deadlock in debug mode it'd be great
>>> if you send me a stack trace of the main thread (not sure if it's
>>> easily possible to copy *all* stack traces from the debugger).
>>>
>>> Cheers
>>> /Eike
>>>
|
|
|
| Re: [CDO] Hanging (client) [message #94488 is a reply to message #94413] |
Tue, 28 August 2007 11:11 |
| Eclipse User |
|
|
|
Originally posted by: stepper.sympedia.de
Ed,
Wow, I'm really impressed ;-)
With the both tools mentioned at the bottom of
http://wiki.eclipse.org/How_to_report_a_deadlock it really works.
This is really a tremendous help in tracking down deadlocks!!!
Many many thanks!!!
/Eike
Ed Merks schrieb:
> Eike,
>
> I read somewhere that you can generate a thread dump after a deadlock
> occurs under Unix/Linux by running kill -QUIT <pid> and under Windows
> by hitting Ctrl + Break.
>
>
> Eike Stepper wrote:
>> Simon McDuff schrieb:
>>> Eike,
>>>
>>> Sometimes the server and the client hang...
>>>
>>> If I stop the client and restart and a new one.. it is working
>>> correctly.
>>>
>>> Each time I try to run it in debug mode... it doesn't hang. I only
>>> see the behavior in release.
>>>
>> That's typical for threading problems (deadlocks).
>> Sometime there's a chance to hit it in debug mode if all tracing is
>> disabled (or sent to file instead of the console).
>>
>>> I wasn't sure at the beginning.. but I have that problem for the
>>> last 2 weeks.
>>>
>>> Have you seen this behavior in you environment ?
>>>
>> Yes, unfortunately I have this phenomenon aproximately once or twice
>> a month which is really rare and most of the times without debug mode
>> ;-(
>> This is one of the most urgent issues to fix (probably in Net4j).
>> Since it occurs so rarely I planned to investigate it after my work
>> with the cdo mapper is at least so far that users can try a
>> horizontal mapping with a table per reference.
>>
>> I can remember of a single place in Net4j where I had the feeling a
>> race condition could occur.
>> When I recall this place I'll tell you about.
>>
>> I know there are some tools available that try to find
>> synchronization problems through static code analysis.
>> Do you have experinece with one of these tools?
>>
>>> I'm using Java sun 1.5_0_10
>>>
>> I'm using a jre1.6.0_02 for my IDE and the code is compiled against a
>> J2SE-1.5 execution environment which is associated with a jdk1.5.0_12
>> here at the moment.
>>
>> If you are ever able to hit the deadlock in debug mode it'd be great
>> if you send me a stack trace of the main thread (not sure if it's
>> easily possible to copy *all* stack traces from the debugger).
>>
>> Cheers
>> /Eike
>>
|
|
|
| Re: [CDO] Hanging (client) [message #94519 is a reply to message #94443] |
Tue, 28 August 2007 11:14 |
| Eclipse User |
|
|
|
Originally posted by: merks.ca.ibm.com
This is a multi-part message in MIME format.
--------------030705030603070201020208
Content-Type: text/plain; charset=ISO-8859-15; format=flowed
Content-Transfer-Encoding: 7bit
Eike,
I run Eclipse like this:
eclipse.exe -consolelog
When I hit ctrl-break in the console it prints
JVMDUMP006I Processing Dump Event "user", detail "" - Please Wait.
JVMDUMP007I JVM Requesting Java Dump using
'd:\sandbox\3.3RC4\eclipse\javacore.20070828.071146.107920.t xt'
JVMDUMP010I Java Dump written to
d:\sandbox\3.3RC4\eclipse\javacore.20070828.071146.107920.tx t
JVMDUMP013I Processed Dump Event "user", detail "".
Too bad if the IDE launched things don't support this. You might open a
bugzilla to ask for it since this is awfully useful.
Eike Stepper wrote:
> Ed,
>
> That sounds really interesting!
> What does "generate" mean? Text in the clipboard? A file?
>
> Or do you mean hitting Ctrl+Break in the concole where the deadlocked
> process is running?
> This at least doesn't work with the IDE console.
>
> Or should I use debug mode?
> No, Then I could simply suspend the locked thread.
>
> Hmmm, it's not clear to me...
> /Eike
>
>
> Ed Merks schrieb:
>> Eike,
>>
>> I read somewhere that you can generate a thread dump after a deadlock
>> occurs under Unix/Linux by running kill -QUIT <pid> and under Windows
>> by hitting Ctrl + Break.
>>
>>
>> Eike Stepper wrote:
>>> Simon McDuff schrieb:
>>>> Eike,
>>>>
>>>> Sometimes the server and the client hang...
>>>>
>>>> If I stop the client and restart and a new one.. it is working
>>>> correctly.
>>>>
>>>> Each time I try to run it in debug mode... it doesn't hang. I only
>>>> see the behavior in release.
>>>>
>>> That's typical for threading problems (deadlocks).
>>> Sometime there's a chance to hit it in debug mode if all tracing is
>>> disabled (or sent to file instead of the console).
>>>
>>>> I wasn't sure at the beginning.. but I have that problem for the
>>>> last 2 weeks.
>>>>
>>>> Have you seen this behavior in you environment ?
>>>>
>>> Yes, unfortunately I have this phenomenon aproximately once or twice
>>> a month which is really rare and most of the times without debug
>>> mode ;-(
>>> This is one of the most urgent issues to fix (probably in Net4j).
>>> Since it occurs so rarely I planned to investigate it after my work
>>> with the cdo mapper is at least so far that users can try a
>>> horizontal mapping with a table per reference.
>>>
>>> I can remember of a single place in Net4j where I had the feeling a
>>> race condition could occur.
>>> When I recall this place I'll tell you about.
>>>
>>> I know there are some tools available that try to find
>>> synchronization problems through static code analysis.
>>> Do you have experinece with one of these tools?
>>>
>>>> I'm using Java sun 1.5_0_10
>>>>
>>> I'm using a jre1.6.0_02 for my IDE and the code is compiled against
>>> a J2SE-1.5 execution environment which is associated with a
>>> jdk1.5.0_12 here at the moment.
>>>
>>> If you are ever able to hit the deadlock in debug mode it'd be great
>>> if you send me a stack trace of the main thread (not sure if it's
>>> easily possible to copy *all* stack traces from the debugger).
>>>
>>> Cheers
>>> /Eike
>>>
--------------030705030603070201020208
Content-Type: text/html; charset=ISO-8859-15
Content-Transfer-Encoding: 8bit
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-15"
http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Eike,<br>
<br>
I run Eclipse like this:<br>
<blockquote>eclipse.exe -consolelog<br>
</blockquote>
When I hit ctrl-break in the console it prints<br>
<blockquote>JVMDUMP006I Processing Dump Event "user", detail "" -
Please Wait.<br>
JVMDUMP007I JVM Requesting Java Dump using
'd:\sandbox\3.3RC4\eclipse\javacore.20070828.071146.107920.t xt' <br>
JVMDUMP010I Java Dump written to
d:\sandbox\3.3RC4\eclipse\javacore.20070828.071146.107920.tx t <br>
JVMDUMP013I Processed Dump Event "user", detail "".<br>
</blockquote>
Too bad if the IDE launched things don't support this.
|
|
|
| Re: [CDO] Hanging (client) [message #94551 is a reply to message #94488] |
Tue, 28 August 2007 11:16 |
| Eclipse User |
|
|
|
Originally posted by: merks.ca.ibm.com
Eike,
I should read all the notes before replying. :-)
Eike Stepper wrote:
> Ed,
>
> Wow, I'm really impressed ;-)
> With the both tools mentioned at the bottom of
> http://wiki.eclipse.org/How_to_report_a_deadlock it really works.
> This is really a tremendous help in tracking down deadlocks!!!
>
> Many many thanks!!!
> /Eike
>
>
> Ed Merks schrieb:
>> Eike,
>>
>> I read somewhere that you can generate a thread dump after a deadlock
>> occurs under Unix/Linux by running kill -QUIT <pid> and under Windows
>> by hitting Ctrl + Break.
>>
>>
>> Eike Stepper wrote:
>>> Simon McDuff schrieb:
>>>> Eike,
>>>>
>>>> Sometimes the server and the client hang...
>>>>
>>>> If I stop the client and restart and a new one.. it is working
>>>> correctly.
>>>>
>>>> Each time I try to run it in debug mode... it doesn't hang. I only
>>>> see the behavior in release.
>>>>
>>> That's typical for threading problems (deadlocks).
>>> Sometime there's a chance to hit it in debug mode if all tracing is
>>> disabled (or sent to file instead of the console).
>>>
>>>> I wasn't sure at the beginning.. but I have that problem for the
>>>> last 2 weeks.
>>>>
>>>> Have you seen this behavior in you environment ?
>>>>
>>> Yes, unfortunately I have this phenomenon aproximately once or twice
>>> a month which is really rare and most of the times without debug
>>> mode ;-(
>>> This is one of the most urgent issues to fix (probably in Net4j).
>>> Since it occurs so rarely I planned to investigate it after my work
>>> with the cdo mapper is at least so far that users can try a
>>> horizontal mapping with a table per reference.
>>>
>>> I can remember of a single place in Net4j where I had the feeling a
>>> race condition could occur.
>>> When I recall this place I'll tell you about.
>>>
>>> I know there are some tools available that try to find
>>> synchronization problems through static code analysis.
>>> Do you have experinece with one of these tools?
>>>
>>>> I'm using Java sun 1.5_0_10
>>>>
>>> I'm using a jre1.6.0_02 for my IDE and the code is compiled against
>>> a J2SE-1.5 execution environment which is associated with a
>>> jdk1.5.0_12 here at the moment.
>>>
>>> If you are ever able to hit the deadlock in debug mode it'd be great
>>> if you send me a stack trace of the main thread (not sure if it's
>>> easily possible to copy *all* stack traces from the debugger).
>>>
>>> Cheers
>>> /Eike
>>>
|
|
|
| Re: [CDO] Hanging (client) [message #94566 is a reply to message #94273] |
Tue, 28 August 2007 11:16 |
| Eclipse User |
|
|
|
Originally posted by: stepper.sympedia.de
Simon,
I guess you have followed the other postings to this thread.
If not, here is a short explanation of what to do if deadlock occurs in
a non-debug runtime launch:
1. Download and run
http://www.microsoft.com/technet/sysinternals/utilities/Proc essExplorer.mspx
2. Find the pid of the launched process under Eclipse.exe
3. Download
http://www.latenighthacking.com/projects/2003/sendSignal/Sen dSignal.exe
4. Run SendSignal.exe <pid> in any console
5. Copy the stack traces from the IDE console and send them to me
Cheers
/Eike
Simon McDuff schrieb:
> Eike,
>
> Sometimes the server and the client hang...
>
> If I stop the client and restart and a new one.. it is working correctly.
>
> Each time I try to run it in debug mode... it doesn't hang. I only see the
> behavior in release.
>
> I wasn't sure at the beginning.. but I have that problem for the last 2
> weeks.
>
> Have you seen this behavior in you environment ?
>
> I'm using Java sun 1.5_0_10
>
>
>
>
>
|
|
|
| Re: [CDO] Hanging (client) [message #94581 is a reply to message #94551] |
Tue, 28 August 2007 11:19 |
| Eclipse User |
|
|
|
Originally posted by: stepper.sympedia.de
Ed Merks schrieb:
> Eike,
>
> I should read all the notes before replying. :-)
....and I should try everything before I spam the group ;-)
Cheers
/Eike
|
|
|
|
| Re: [CDO] Hanging (client) [message #609547 is a reply to message #94273] |
Tue, 28 August 2007 05:45 |
|
Simon McDuff schrieb:
> Eike,
>
> Sometimes the server and the client hang...
>
> If I stop the client and restart and a new one.. it is working correctly.
>
> Each time I try to run it in debug mode... it doesn't hang. I only see the
> behavior in release.
>
That's typical for threading problems (deadlocks).
Sometime there's a chance to hit it in debug mode if all tracing is
disabled (or sent to file instead of the console).
> I wasn't sure at the beginning.. but I have that problem for the last 2
> weeks.
>
> Have you seen this behavior in you environment ?
>
Yes, unfortunately I have this phenomenon aproximately once or twice a
month which is really rare and most of the times without debug mode ;-(
This is one of the most urgent issues to fix (probably in Net4j). Since
it occurs so rarely I planned to investigate it after my work with the
cdo mapper is at least so far that users can try a horizontal mapping
with a table per reference.
I can remember of a single place in Net4j where I had the feeling a race
condition could occur.
When I recall this place I'll tell you about.
I know there are some tools available that try to find synchronization
problems through static code analysis.
Do you have experinece with one of these tools?
> I'm using Java sun 1.5_0_10
>
I'm using a jre1.6.0_02 for my IDE and the code is compiled against a
J2SE-1.5 execution environment which is associated with a jdk1.5.0_12
here at the moment.
If you are ever able to hit the deadlock in debug mode it'd be great if
you send me a stack trace of the main thread (not sure if it's easily
possible to copy *all* stack traces from the debugger).
Cheers
/Eike
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
| Re: [CDO] Hanging (client) [message #609550 is a reply to message #94370] |
Tue, 28 August 2007 09:53 |
 Ed Merks
Messages: 33133
Registered: July 2009 |
Senior Member |
|
|
Eike,
I read somewhere that you can generate a thread dump after a deadlock
occurs under Unix/Linux by running kill -QUIT <pid> and under Windows by
hitting Ctrl + Break.
Eike Stepper wrote:
> Simon McDuff schrieb:
>> Eike,
>>
>> Sometimes the server and the client hang...
>>
>> If I stop the client and restart and a new one.. it is working
>> correctly.
>>
>> Each time I try to run it in debug mode... it doesn't hang. I only
>> see the behavior in release.
>>
> That's typical for threading problems (deadlocks).
> Sometime there's a chance to hit it in debug mode if all tracing is
> disabled (or sent to file instead of the console).
>
>> I wasn't sure at the beginning.. but I have that problem for the last
>> 2 weeks.
>>
>> Have you seen this behavior in you environment ?
>>
> Yes, unfortunately I have this phenomenon aproximately once or twice a
> month which is really rare and most of the times without debug mode ;-(
> This is one of the most urgent issues to fix (probably in Net4j).
> Since it occurs so rarely I planned to investigate it after my work
> with the cdo mapper is at least so far that users can try a horizontal
> mapping with a table per reference.
>
> I can remember of a single place in Net4j where I had the feeling a
> race condition could occur.
> When I recall this place I'll tell you about.
>
> I know there are some tools available that try to find synchronization
> problems through static code analysis.
> Do you have experinece with one of these tools?
>
>> I'm using Java sun 1.5_0_10
>>
> I'm using a jre1.6.0_02 for my IDE and the code is compiled against a
> J2SE-1.5 execution environment which is associated with a jdk1.5.0_12
> here at the moment.
>
> If you are ever able to hit the deadlock in debug mode it'd be great
> if you send me a stack trace of the main thread (not sure if it's
> easily possible to copy *all* stack traces from the debugger).
>
> Cheers
> /Eike
>
Ed Merks
Professional Support: https://www.macromodeling.com/
|
|
|
| Re: [CDO] Hanging (client) [message #609552 is a reply to message #94413] |
Tue, 28 August 2007 10:36 |
|
Ed,
That sounds really interesting!
What does "generate" mean? Text in the clipboard? A file?
Or do you mean hitting Ctrl+Break in the concole where the deadlocked
process is running?
This at least doesn't work with the IDE console.
Or should I use debug mode?
No, Then I could simply suspend the locked thread.
Hmmm, it's not clear to me...
/Eike
Ed Merks schrieb:
> Eike,
>
> I read somewhere that you can generate a thread dump after a deadlock
> occurs under Unix/Linux by running kill -QUIT <pid> and under Windows
> by hitting Ctrl + Break.
>
>
> Eike Stepper wrote:
>> Simon McDuff schrieb:
>>> Eike,
>>>
>>> Sometimes the server and the client hang...
>>>
>>> If I stop the client and restart and a new one.. it is working
>>> correctly.
>>>
>>> Each time I try to run it in debug mode... it doesn't hang. I only
>>> see the behavior in release.
>>>
>> That's typical for threading problems (deadlocks).
>> Sometime there's a chance to hit it in debug mode if all tracing is
>> disabled (or sent to file instead of the console).
>>
>>> I wasn't sure at the beginning.. but I have that problem for the
>>> last 2 weeks.
>>>
>>> Have you seen this behavior in you environment ?
>>>
>> Yes, unfortunately I have this phenomenon aproximately once or twice
>> a month which is really rare and most of the times without debug mode
>> ;-(
>> This is one of the most urgent issues to fix (probably in Net4j).
>> Since it occurs so rarely I planned to investigate it after my work
>> with the cdo mapper is at least so far that users can try a
>> horizontal mapping with a table per reference.
>>
>> I can remember of a single place in Net4j where I had the feeling a
>> race condition could occur.
>> When I recall this place I'll tell you about.
>>
>> I know there are some tools available that try to find
>> synchronization problems through static code analysis.
>> Do you have experinece with one of these tools?
>>
>>> I'm using Java sun 1.5_0_10
>>>
>> I'm using a jre1.6.0_02 for my IDE and the code is compiled against a
>> J2SE-1.5 execution environment which is associated with a jdk1.5.0_12
>> here at the moment.
>>
>> If you are ever able to hit the deadlock in debug mode it'd be great
>> if you send me a stack trace of the main thread (not sure if it's
>> easily possible to copy *all* stack traces from the debugger).
>>
>> Cheers
>> /Eike
>>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
| Re: [CDO] Hanging (client) [message #609553 is a reply to message #94443] |
Tue, 28 August 2007 10:41 |
|
http://java.sun.com/developer/technicalArticles/Programming/ Stacktrace/
;-)
Eike Stepper schrieb:
> Ed,
>
> That sounds really interesting!
> What does "generate" mean? Text in the clipboard? A file?
>
> Or do you mean hitting Ctrl+Break in the concole where the deadlocked
> process is running?
> This at least doesn't work with the IDE console.
>
> Or should I use debug mode?
> No, Then I could simply suspend the locked thread.
>
> Hmmm, it's not clear to me...
> /Eike
>
>
> Ed Merks schrieb:
>> Eike,
>>
>> I read somewhere that you can generate a thread dump after a deadlock
>> occurs under Unix/Linux by running kill -QUIT <pid> and under Windows
>> by hitting Ctrl + Break.
>>
>>
>> Eike Stepper wrote:
>>> Simon McDuff schrieb:
>>>> Eike,
>>>>
>>>> Sometimes the server and the client hang...
>>>>
>>>> If I stop the client and restart and a new one.. it is working
>>>> correctly.
>>>>
>>>> Each time I try to run it in debug mode... it doesn't hang. I only
>>>> see the behavior in release.
>>>>
>>> That's typical for threading problems (deadlocks).
>>> Sometime there's a chance to hit it in debug mode if all tracing is
>>> disabled (or sent to file instead of the console).
>>>
>>>> I wasn't sure at the beginning.. but I have that problem for the
>>>> last 2 weeks.
>>>>
>>>> Have you seen this behavior in you environment ?
>>>>
>>> Yes, unfortunately I have this phenomenon aproximately once or twice
>>> a month which is really rare and most of the times without debug
>>> mode ;-(
>>> This is one of the most urgent issues to fix (probably in Net4j).
>>> Since it occurs so rarely I planned to investigate it after my work
>>> with the cdo mapper is at least so far that users can try a
>>> horizontal mapping with a table per reference.
>>>
>>> I can remember of a single place in Net4j where I had the feeling a
>>> race condition could occur.
>>> When I recall this place I'll tell you about.
>>>
>>> I know there are some tools available that try to find
>>> synchronization problems through static code analysis.
>>> Do you have experinece with one of these tools?
>>>
>>>> I'm using Java sun 1.5_0_10
>>>>
>>> I'm using a jre1.6.0_02 for my IDE and the code is compiled against
>>> a J2SE-1.5 execution environment which is associated with a
>>> jdk1.5.0_12 here at the moment.
>>>
>>> If you are ever able to hit the deadlock in debug mode it'd be great
>>> if you send me a stack trace of the main thread (not sure if it's
>>> easily possible to copy *all* stack traces from the debugger).
>>>
>>> Cheers
>>> /Eike
>>>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
| Re: [CDO] Hanging (client) [message #609554 is a reply to message #94443] |
Tue, 28 August 2007 10:53 |
|
This one was also useful:
http://wiki.eclipse.org/How_to_report_a_deadlock
Eike Stepper schrieb:
> Ed,
>
> That sounds really interesting!
> What does "generate" mean? Text in the clipboard? A file?
>
> Or do you mean hitting Ctrl+Break in the concole where the deadlocked
> process is running?
> This at least doesn't work with the IDE console.
>
> Or should I use debug mode?
> No, Then I could simply suspend the locked thread.
>
> Hmmm, it's not clear to me...
> /Eike
>
>
> Ed Merks schrieb:
>> Eike,
>>
>> I read somewhere that you can generate a thread dump after a deadlock
>> occurs under Unix/Linux by running kill -QUIT <pid> and under Windows
>> by hitting Ctrl + Break.
>>
>>
>> Eike Stepper wrote:
>>> Simon McDuff schrieb:
>>>> Eike,
>>>>
>>>> Sometimes the server and the client hang...
>>>>
>>>> If I stop the client and restart and a new one.. it is working
>>>> correctly.
>>>>
>>>> Each time I try to run it in debug mode... it doesn't hang. I only
>>>> see the behavior in release.
>>>>
>>> That's typical for threading problems (deadlocks).
>>> Sometime there's a chance to hit it in debug mode if all tracing is
>>> disabled (or sent to file instead of the console).
>>>
>>>> I wasn't sure at the beginning.. but I have that problem for the
>>>> last 2 weeks.
>>>>
>>>> Have you seen this behavior in you environment ?
>>>>
>>> Yes, unfortunately I have this phenomenon aproximately once or twice
>>> a month which is really rare and most of the times without debug
>>> mode ;-(
>>> This is one of the most urgent issues to fix (probably in Net4j).
>>> Since it occurs so rarely I planned to investigate it after my work
>>> with the cdo mapper is at least so far that users can try a
>>> horizontal mapping with a table per reference.
>>>
>>> I can remember of a single place in Net4j where I had the feeling a
>>> race condition could occur.
>>> When I recall this place I'll tell you about.
>>>
>>> I know there are some tools available that try to find
>>> synchronization problems through static code analysis.
>>> Do you have experinece with one of these tools?
>>>
>>>> I'm using Java sun 1.5_0_10
>>>>
>>> I'm using a jre1.6.0_02 for my IDE and the code is compiled against
>>> a J2SE-1.5 execution environment which is associated with a
>>> jdk1.5.0_12 here at the moment.
>>>
>>> If you are ever able to hit the deadlock in debug mode it'd be great
>>> if you send me a stack trace of the main thread (not sure if it's
>>> easily possible to copy *all* stack traces from the debugger).
>>>
>>> Cheers
>>> /Eike
>>>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
| Re: [CDO] Hanging (client) [message #609555 is a reply to message #94413] |
Tue, 28 August 2007 11:11 |
|
Ed,
Wow, I'm really impressed ;-)
With the both tools mentioned at the bottom of
http://wiki.eclipse.org/How_to_report_a_deadlock it really works.
This is really a tremendous help in tracking down deadlocks!!!
Many many thanks!!!
/Eike
Ed Merks schrieb:
> Eike,
>
> I read somewhere that you can generate a thread dump after a deadlock
> occurs under Unix/Linux by running kill -QUIT <pid> and under Windows
> by hitting Ctrl + Break.
>
>
> Eike Stepper wrote:
>> Simon McDuff schrieb:
>>> Eike,
>>>
>>> Sometimes the server and the client hang...
>>>
>>> If I stop the client and restart and a new one.. it is working
>>> correctly.
>>>
>>> Each time I try to run it in debug mode... it doesn't hang. I only
>>> see the behavior in release.
>>>
>> That's typical for threading problems (deadlocks).
>> Sometime there's a chance to hit it in debug mode if all tracing is
>> disabled (or sent to file instead of the console).
>>
>>> I wasn't sure at the beginning.. but I have that problem for the
>>> last 2 weeks.
>>>
>>> Have you seen this behavior in you environment ?
>>>
>> Yes, unfortunately I have this phenomenon aproximately once or twice
>> a month which is really rare and most of the times without debug mode
>> ;-(
>> This is one of the most urgent issues to fix (probably in Net4j).
>> Since it occurs so rarely I planned to investigate it after my work
>> with the cdo mapper is at least so far that users can try a
>> horizontal mapping with a table per reference.
>>
>> I can remember of a single place in Net4j where I had the feeling a
>> race condition could occur.
>> When I recall this place I'll tell you about.
>>
>> I know there are some tools available that try to find
>> synchronization problems through static code analysis.
>> Do you have experinece with one of these tools?
>>
>>> I'm using Java sun 1.5_0_10
>>>
>> I'm using a jre1.6.0_02 for my IDE and the code is compiled against a
>> J2SE-1.5 execution environment which is associated with a jdk1.5.0_12
>> here at the moment.
>>
>> If you are ever able to hit the deadlock in debug mode it'd be great
>> if you send me a stack trace of the main thread (not sure if it's
>> easily possible to copy *all* stack traces from the debugger).
>>
>> Cheers
>> /Eike
>>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
| Re: [CDO] Hanging (client) [message #609557 is a reply to message #94443] |
Tue, 28 August 2007 11:14 |
Ed Merks
Messages: 33133
Registered: July 2009 |
Senior Member |
|
|
This is a multi-part message in MIME format.
--------------030705030603070201020208
Content-Type: text/plain; charset=ISO-8859-15; format=flowed
Content-Transfer-Encoding: 7bit
Eike,
I run Eclipse like this:
eclipse.exe -consolelog
When I hit ctrl-break in the console it prints
JVMDUMP006I Processing Dump Event "user", detail "" - Please Wait.
JVMDUMP007I JVM Requesting Java Dump using
'd:\sandbox\3.3RC4\eclipse\javacore.20070828.071146.107920.t xt'
JVMDUMP010I Java Dump written to
d:\sandbox\3.3RC4\eclipse\javacore.20070828.071146.107920.tx t
JVMDUMP013I Processed Dump Event "user", detail "".
Too bad if the IDE launched things don't support this. You might open a
bugzilla to ask for it since this is awfully useful.
Eike Stepper wrote:
> Ed,
>
> That sounds really interesting!
> What does "generate" mean? Text in the clipboard? A file?
>
> Or do you mean hitting Ctrl+Break in the concole where the deadlocked
> process is running?
> This at least doesn't work with the IDE console.
>
> Or should I use debug mode?
> No, Then I could simply suspend the locked thread.
>
> Hmmm, it's not clear to me...
> /Eike
>
>
> Ed Merks schrieb:
>> Eike,
>>
>> I read somewhere that you can generate a thread dump after a deadlock
>> occurs under Unix/Linux by running kill -QUIT <pid> and under Windows
>> by hitting Ctrl + Break.
>>
>>
>> Eike Stepper wrote:
>>> Simon McDuff schrieb:
>>>> Eike,
>>>>
>>>> Sometimes the server and the client hang...
>>>>
>>>> If I stop the client and restart and a new one.. it is working
>>>> correctly.
>>>>
>>>> Each time I try to run it in debug mode... it doesn't hang. I only
>>>> see the behavior in release.
>>>>
>>> That's typical for threading problems (deadlocks).
>>> Sometime there's a chance to hit it in debug mode if all tracing is
>>> disabled (or sent to file instead of the console).
>>>
>>>> I wasn't sure at the beginning.. but I have that problem for the
>>>> last 2 weeks.
>>>>
>>>> Have you seen this behavior in you environment ?
>>>>
>>> Yes, unfortunately I have this phenomenon aproximately once or twice
>>> a month which is really rare and most of the times without debug
>>> mode ;-(
>>> This is one of the most urgent issues to fix (probably in Net4j).
>>> Since it occurs so rarely I planned to investigate it after my work
>>> with the cdo mapper is at least so far that users can try a
>>> horizontal mapping with a table per reference.
>>>
>>> I can remember of a single place in Net4j where I had the feeling a
>>> race condition could occur.
>>> When I recall this place I'll tell you about.
>>>
>>> I know there are some tools available that try to find
>>> synchronization problems through static code analysis.
>>> Do you have experinece with one of these tools?
>>>
>>>> I'm using Java sun 1.5_0_10
>>>>
>>> I'm using a jre1.6.0_02 for my IDE and the code is compiled against
>>> a J2SE-1.5 execution environment which is associated with a
>>> jdk1.5.0_12 here at the moment.
>>>
>>> If you are ever able to hit the deadlock in debug mode it'd be great
>>> if you send me a stack trace of the main thread (not sure if it's
>>> easily possible to copy *all* stack traces from the debugger).
>>>
>>> Cheers
>>> /Eike
>>>
--------------030705030603070201020208
Content-Type: text/html; charset=ISO-8859-15
Content-Transfer-Encoding: 8bit
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-15"
http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Eike,<br>
<br>
I run Eclipse like this:<br>
<blockquote>eclipse.exe -consolelog<br>
</blockquote>
When I hit ctrl-break in the console it prints<br>
<blockquote>JVMDUMP006I Processing Dump Event "user", detail "" -
Please Wait.<br>
JVMDUMP007I JVM Requesting Java Dump using
'd:\sandbox\3.3RC4\eclipse\javacore.20070828.071146.107920.t xt' <br>
JVMDUMP010I Java Dump written to
d:\sandbox\3.3RC4\eclipse\javacore.20070828.071146.107920.tx t <br>
JVMDUMP013I Processed Dump Event "user", detail "".<br>
</blockquote>
Too bad if the IDE launched things don't support this.
Ed Merks
Professional Support: https://www.macromodeling.com/
|
|
|
| Re: [CDO] Hanging (client) [message #609559 is a reply to message #94488] |
Tue, 28 August 2007 11:16 |
Ed Merks
Messages: 33133
Registered: July 2009 |
Senior Member |
|
|
Eike,
I should read all the notes before replying. :-)
Eike Stepper wrote:
> Ed,
>
> Wow, I'm really impressed ;-)
> With the both tools mentioned at the bottom of
> http://wiki.eclipse.org/How_to_report_a_deadlock it really works.
> This is really a tremendous help in tracking down deadlocks!!!
>
> Many many thanks!!!
> /Eike
>
>
> Ed Merks schrieb:
>> Eike,
>>
>> I read somewhere that you can generate a thread dump after a deadlock
>> occurs under Unix/Linux by running kill -QUIT <pid> and under Windows
>> by hitting Ctrl + Break.
>>
>>
>> Eike Stepper wrote:
>>> Simon McDuff schrieb:
>>>> Eike,
>>>>
>>>> Sometimes the server and the client hang...
>>>>
>>>> If I stop the client and restart and a new one.. it is working
>>>> correctly.
>>>>
>>>> Each time I try to run it in debug mode... it doesn't hang. I only
>>>> see the behavior in release.
>>>>
>>> That's typical for threading problems (deadlocks).
>>> Sometime there's a chance to hit it in debug mode if all tracing is
>>> disabled (or sent to file instead of the console).
>>>
>>>> I wasn't sure at the beginning.. but I have that problem for the
>>>> last 2 weeks.
>>>>
>>>> Have you seen this behavior in you environment ?
>>>>
>>> Yes, unfortunately I have this phenomenon aproximately once or twice
>>> a month which is really rare and most of the times without debug
>>> mode ;-(
>>> This is one of the most urgent issues to fix (probably in Net4j).
>>> Since it occurs so rarely I planned to investigate it after my work
>>> with the cdo mapper is at least so far that users can try a
>>> horizontal mapping with a table per reference.
>>>
>>> I can remember of a single place in Net4j where I had the feeling a
>>> race condition could occur.
>>> When I recall this place I'll tell you about.
>>>
>>> I know there are some tools available that try to find
>>> synchronization problems through static code analysis.
>>> Do you have experinece with one of these tools?
>>>
>>>> I'm using Java sun 1.5_0_10
>>>>
>>> I'm using a jre1.6.0_02 for my IDE and the code is compiled against
>>> a J2SE-1.5 execution environment which is associated with a
>>> jdk1.5.0_12 here at the moment.
>>>
>>> If you are ever able to hit the deadlock in debug mode it'd be great
>>> if you send me a stack trace of the main thread (not sure if it's
>>> easily possible to copy *all* stack traces from the debugger).
>>>
>>> Cheers
>>> /Eike
>>>
Ed Merks
Professional Support: https://www.macromodeling.com/
|
|
|
|
|
|
] 
Search
Help
Register
Login
Home
