Java crashes - only message is "Aborted"???

Thread view:

dduck - 27 Jun 2007 14:37 GMT

Hi,

We have a strange problem. Our program occationally crashes after
running for several hours. Nothing is logged, and the only message
written to the console is "Aborted" - no stack trace, no exception, no
nothing. We have encapsulated the invocation in try/catch Throwable in
an attempt to at least be able to log a stack trace, but apparently no
exception is thrown. No core dump or fatal error log is produced.

Our code leverages a Java SDK from Canto (Cumulus), which uses native
code via JNI invocation. We are currently running the program with
option verbose:jni, in the hope that this will help us pinpoint the
error, and are currently eagerly awaiting a crash.

Any suggestions on how to proceed?

Sincerely,
Anders S. Johansen

Reply to this Message

Gordon Beaton - 27 Jun 2007 14:54 GMT

> We have a strange problem. Our program occationally crashes after
> running for several hours. Nothing is logged, and the only message
> written to the console is "Aborted" - no stack trace, no exception,
> no nothing.

[...]

> No core dump or fatal error log is produced.

[...]

> Any suggestions on how to proceed?

You've received SIGABRT, either because someone has sent that signal
(with kill), or because the code itself invokes abort(), perhaps via
assert().

Try doing "ulimit -c unlimited" in the shell (or start script) before
starting the application, or confirm with "ulimit -c" that the setting
is "unlimited". That should get you a corefile.

/gordon

--

Reply to this Message

dduck - 27 Jun 2007 15:13 GMT

> Try doing "ulimit -c unlimited" in the shell (or start script) before
> starting the application, or confirm with "ulimit -c" that the setting
> is "unlimited". That should get you a corefile.

Thank you. We will try that.

Sincerely,
Anders

Reply to this Message

Nigel Wade - 27 Jun 2007 17:07 GMT

>> Try doing "ulimit -c unlimited" in the shell (or start script) before
>> starting the application, or confirm with "ulimit -c" that the setting
[quoted text clipped - 4 lines]
> Sincerely,
> Anders

What OS are you running this on?

You might be hitting a resource limit which is aborting the command. What limits
are imposed (CPU, stack size, vm size etc) in your environment?

Signature

Nigel Wade, System Administrator, Space Plasma Physics Group,
University of Leicester, Leicester, LE1 7RH, UK
E-mail : nmw@ion.le.ac.uk
Phone : +44 (0)116 2523548, Fax : +44 (0)116 2523555

Reply to this Message

dduck - 27 Jun 2007 18:14 GMT

> >> Try doing "ulimit -c unlimited" in the shell (or start script) before
> >> starting the application, or confirm with "ulimit -c" that the setting
[quoted text clipped - 9 lines]
> You might be hitting a resource limit which is aborting the command. What limits
> are imposed (CPU, stack size, vm size etc) in your environment?

We are running on RedHat Linux on an x86 arch. I will post precise
info on this tomorrow when I get back to the office (no remote access
to that particular machine).

We constantly monitor the app with JConsole. There are no indications
that any obvious resource is being exhausted (CPU, memory, threads,
classes loaded...).

Thank you for your suggestions - I will look into it and report the
findings.

Sincerely,
Anders

Reply to this Message

Nigel Wade - 28 Jun 2007 09:29 GMT

>> >> Try doing "ulimit -c unlimited" in the shell (or start script) before
>> >> starting the application, or confirm with "ulimit -c" that the setting
[quoted text clipped - 23 lines]
> Sincerely,
> Anders

You won't get any indication that a resource limit is nearly exhausted, your
program only receives the indication when it is exhausted. That may just be a
failure to allocate a new resource, such as failing to open a file because you
already have open the max. allowed, or the program being aborted due to
exceeding the CPU limit.

"ulimit -a" should tell you your limits in Linux.

Signature

Nigel Wade, System Administrator, Space Plasma Physics Group,
University of Leicester, Leicester, LE1 7RH, UK
E-mail : nmw@ion.le.ac.uk
Phone : +44 (0)116 2523548, Fax : +44 (0)116 2523555

Reply to this Message

Roedy Green - 28 Jun 2007 15:04 GMT

> the only message
>written to the console is "Aborted"

do a global search of your source code for the string "Aborted". It
may be the app itself exiting on some "impossible" condition.
--
Roedy Green Canadian Mind Products
The Java Glossary
http://mindprod.com

Reply to this Message

David Gourley - 28 Jun 2007 21:03 GMT

>>the only message
>>written to the console is "Aborted"
[quoted text clipped - 5 lines]
> The Java Glossary
> http://mindprod.com

Or if that doesn't show up, truss (or whatever the Linux equivalent is)
the JVM running your application: this will tell you what system call
failed with an errno which you can look up in the man pages - this might
give a clue which resource got exhausted. I think the compile thread
used to abort if it can't allocate memory...

Dave

Reply to this Message

dduck - 10 Aug 2007 06:52 GMT

Hi all,

Just for the record: The problem i solved, and it was indeed resource
allocation innon-java code that was the culprit. It turned out that
the Canto code under certain circumstances would not perform its
internal garbage collection, which eventuelly led to memory usage
hitting the hardware wall at 2GB.

As all memory was allocated, there was of course no java core file,
and no stack trace. I'm still trying to figure out why there was no
ordinary core file, but that's mainly to satisfy my curiosity.

Thank you for your help - it was really useful!

Sincerely,
Anders S. Johansen, Royal Danish Library

> We have a strange problem. Our program occationally crashes after
> running for several hours. Nothing is logged, and the only message
[quoted text clipped - 7 lines]
> option verbose:jni, in the hope that this will help us pinpoint the
> error, and are currently eagerly awaiting a crash.

Reply to this Message

Java crashes - only message is "Aborted"???

Free Magazines