[Pljava-dev] pljava crash in jre

test platform is: centos6 minimal install, with or without the
almost-6.1 'cr' files. 64bit. jdk6u29, pgsql 9.0.5 or 9.0.6...
postgresql-9.0 installed from depesz' yum.postgresql.org repository.
pljava binaries built exactly how I detailed here a week back, running
in /usr/share/pljava/ .

so I've been fighting this hard crash as soon as I ran any pljava on our
dev servers. crashes as soon as you run any pljava, including the
crash-n-burn test: select sysprop( 'java.home' ); nothing I tried
would work...

crash leaves this lengthy debug file in the /var/lib/pgsql/9.0/data
directory
http://freescruz.com/hs_err_pid1704.log

it was crashing like this both on a 'real' 12 core 48gb server, and on a
4-core 6gb VMware ESXi instance, with the same basic configuration. I
tried all sorts of variations, still hard crashes as soon as you run any
pljava, even sysprop

I built a small VM at home, and installed the same versions of
everything... and it worked. :-O

narrowed it down. 1 or 4 CPU cores doesn't matter. our 'large' .conf
file crashes, and a default small file works. {!!!},

I've narrowed it down to...

shared_buffers=2048MB crashes,
shared_buffers=1024MB works.

--
john r pierce N 37, W 122
santa cruz ca mid-left coast

On 12/08/11 11:13 PM, John R Pierce wrote:
>
> I've narrowed it down to...
>
> shared_buffers=2048MB crashes,
> shared_buffers=1024MB works.

1700MB works. 1800MB crashes. 1750MB works.

and, per Johann's suggestion, I tried pointing /usr/java/latest to
OpenJDK...

piercej=# select sysprop( 'java.home' );
sysprop
----------------------------------------------------
/usr/lib/jvm/java-1.6.0-openjdk-1.6.0.0.x86_64/jre
(1 row)

and indeed, same thing. shared_buffers = 1750MB works,
shared_buffers=2048MB crashes.

#
# A fatal error has been detected by the Java Runtime Environment:
#
# SIGSEGV (0xb) at pc=0x00007f4bd693d711, pid=3801, tid=139966157567936
#
# JRE version: 6.0_20-b20
# Java VM: OpenJDK 64-Bit Server VM (19.0-b09 mixed mode linux-amd64
compressed oops)
# Derivative: IcedTea6 1.9.10
# Distribution: CentOS Linux release 6.0 (Final), package
rhel-1.40.1.9.10.el6_1-x86_64
# Problematic frame:
# J java.lang.String.hashCode()I
#
# If you would like to submit a bug report, please include
# instructions how to reproduce the bug and visit:
# http://icedtea.classpath.org/bugzilla
#
....(long crashdump left in hs_err_pidXXXX.log file)

--
john r pierce N 37, W 122
santa cruz ca mid-left coast

On 12/09/11 1:43 PM, John R Pierce wrote:
> On 12/08/11 11:13 PM, John R Pierce wrote:
>>
>> I've narrowed it down to...
>>
>> shared_buffers=2048MB crashes,
>> shared_buffers=1024MB works.
>
> 1700MB works. 1800MB crashes. 1750MB works.

just curious if there's been any update on this?

--
john r pierce N 37, W 122
santa cruz ca mid-left coast

John R Pierce <pierce at hogranch.com> writes:

> On 12/09/11 1:43 PM, John R Pierce wrote:
>> On 12/08/11 11:13 PM, John R Pierce wrote:
>>>
>>> I've narrowed it down to...
>>>
>>> shared_buffers=2048MB crashes,
>>> shared_buffers=1024MB works.
>>
>> 1700MB works. 1800MB crashes. 1750MB works.
>
> just curious if there's been any update on this?

I haven't had time to look into this yet. I suspect this may be a bug
in the Virtual Machine itself. Are you in any position to test with
IBM Java too?

--
Johann Oskarsson http://www.2ndquadrant.com/ |[]
PostgreSQL Development, 24x7 Support, Training and Services --+--
|
Blog: http://my.opera.com/myrkraverk/blog/

On 01/05/12 10:53 PM, Johann 'Myrkraverk' Oskarsson wrote:
> John R Pierce<pierce at hogranch.com> writes:
>
>> > On 12/09/11 1:43 PM, John R Pierce wrote:
>>> >> On 12/08/11 11:13 PM, John R Pierce wrote:
>>>> >>>
>>>> >>> I've narrowed it down to...
>>>> >>>
>>>> >>> shared_buffers=2048MB crashes,
>>>> >>> shared_buffers=1024MB works.
>>> >>
>>> >> 1700MB works. 1800MB crashes. 1750MB works.
>> >
>> > just curious if there's been any update on this?
> I haven't had time to look into this yet. I suspect this may be a bug
> in the Virtual Machine itself. Are you in any position to test with
> IBM Java too?

hmm. In my lab at work, I have IBM Java on a IBM Power server with
AIX... I guess I could build a version for that.

I believe I tested this with OpenJDK as well as Sun JDK and got the same
results.

--
john r pierce N 37, W 122
santa cruz ca mid-left coast

John R Pierce <pierce at hogranch.com> writes:

> On 01/05/12 10:53 PM, Johann 'Myrkraverk' Oskarsson wrote:
> hmm. In my lab at work, I have IBM Java on a IBM Power server with
> AIX... I guess I could build a version for that.
>
> I believe I tested this with OpenJDK as well as Sun JDK and got the
> same results.

You may not need to build a separate version. If you can install IBM
Java for Linux somewhere you may be able to try it with
LD_LIBRARY_PATH or LD_PRELOAD.

--
Johann Oskarsson http://www.2ndquadrant.com/ |[]
PostgreSQL Development, 24x7 Support, Training and Services --+--
|
Blog: http://my.opera.com/myrkraverk/blog/

We hit the same issue....

I loaded IBM JAVA6 and re-pointed by changing /etc/ld.so.conf.d/java.conf to
point to libjvm.
This issue does not occur with IBM JAVA (for us anyway)
I can switch between java.conf files - and only get the issue when using SUN
JAVA.
I tried the latest Java 6 update from SUN and still get the same issue.

--
View this message in context: http://postgresql.1045698.n5.nabble.com/pljava-crash-in-jre-tp5060898p5430580.html
Sent from the PL/Java mailing list archive at Nabble.com.

wilsoncr74 <craig.wilson at hssnet.com> writes:

> We hit the same issue....
>
> I loaded IBM JAVA6 and re-pointed by changing
> /etc/ld.so.conf.d/java.conf to point to libjvm. This issue does not
> occur with IBM JAVA (for us anyway) I can switch between java.conf
> files - and only get the issue when using SUN JAVA. I tried the
> latest Java 6 update from SUN and still get the same issue.

Unfortunately, I do not have direct access to a machine with enough
RAM to test this myself.

It smells of a bug in Sun Java, which is outside the scope of PL/Java
development. That said, if a kind soul is willing to give me shell
access to a machine where I am able to use debug versions of OpenJDK,
PostgreSQL and PL/Java I'm willing to take a look.

--
Johann Oskarsson http://www.2ndquadrant.com/ |[]
PostgreSQL Development, 24x7 Support, Training and Services --+--
|
Blog: http://my.opera.com/myrkraverk/blog/

The environment I used for testing this will be available next week (after
some internal company testing has taken place).

I've had approval to give you access next week if this is ok.

I can discuss connectivity details outside of this forum if this sounds ok.

--
View this message in context: http://postgresql.1045698.n5.nabble.com/pljava-crash-in-jre-tp5060898p5432585.html
Sent from the PL/Java mailing list archive at Nabble.com.

wilsoncr74 <craig.wilson at hssnet.com> writes:

> The environment I used for testing this will be available next week
> (after some internal company testing has taken place).
>
> I've had approval to give you access next week if this is ok.
>
> I can discuss connectivity details outside of this forum if this
> sounds ok.

Yes, I believe I will have some time next week.

I will need a complete development environment; everything needed to
build Postgres, PL/Java and OpenJDK. Already having debug builds
would be great.

Building these is not an issue as long as I don't have to jump through
hoops to do so. Part of the OpenJDK build process is to download
dependencies so blocking network traffic is impolite.

I hope to run the process in a debugger and determine whether the
problem lies within PL/Java or the JVM itself. After all, Pierce
claims this happens with both Sun Java and OpenJDK, and you that it
does not with IBM Java.

If this turns out to be too complex and I find myself debugging the
JVM I'll stop without a support contract.

--
Johann Oskarsson http://www.2ndquadrant.com/ |[]
PostgreSQL Development, 24x7 Support, Training and Services --+--
|
Blog: http://my.opera.com/myrkraverk/blog/