[Opensource] DBMaint crashes the server

[Michael Rimov]

Rick,

Thanks for piping up here.

At 04:45 AM 4/7/2003 -0400, you wrote:
>I saw the same thing a couple of days ago.  Expresso goes into a loop and
>it appears that there is a memory leak.
>
>[Mike asked...]
> > Can you tell us:
> >
> > 1 - How large is the result set being returned?
>
>I was doing wildcard searches, usually returning one record.  The source
>table has about 500 records.

Ok, so we're dealing with small tables, small result sets.  That narrows 
things down considerably.



> > 2 - What database and Driver are you using?
>
>MySQL.  mm.mysql-2.0.14-bin.jar.
>
> > 3 - What JVM/OS?
>
>Sun 1.4.1_01 on Windows 2000 Pro.
>
> > 4 - What version of Tomcat
>
>4.0.6 with Expresso 5.0.3 - a jar that I built.  Browser hangs.  I would
>kill the java process to get my machine back ;)
>
>I should run this with debugging (i.e., logging) turned on and see what
>we can see.  Also, I can run it within a debugger but it *is* intermittant --
>it might take several minutes of work to cause it to happen.

I think this would be worthwhile.  I BELIEVE this is specific to JDK 
1.4.1_01 [not necessarily Sun's VM only]... I've now most likely seen it 
happen on Linux and Mac (Jaguar).... and your report includes Windows VM too.

I first tried running wildcard DBMaint searches in OptimizeIt, and could 
not find anything specific that would be a memory leak.  Then I tried 
running about 100 searches on DBMaint manually on a 575 record 
DefaultUserInfo table, searching on UserName... haven't been able to get 
things to crash. :(  But then again, my Debugger is running JDK 1.3, and 
it's possible that there is something we're doing that JDK 1.4.1_01 doesn't 
like .

I don't think it is the MySQL drivers since I've seen this lock happen once 
using SapDB drivers on a Mac VM.

Can you do me a huge favor, run until a lock and if the debugger can pause, 
tell me where things seem to be locking up in the stacktrace... OR if you 
lose control of your debugger JVM, tell me the last couple of debug entries 
in the log... perhaps that will give us a clue.

Thanks in advance!
                                                 -Mike