You really should try profiling the query and seeing exactly what objects are being allocated. I've used OptimizeIt for this kind of thing and it's amazing the sort of memory leaks you can find. Just stuff you never thought of before.
Also you don't mention if your solution is pure java or not. If you're using any JNI, then look at that code VERY closely. Badly behaved JNI can cause a JVM to just exit out, without any messages or any vaguely helpful info.
If you're getting a thread dump, then maybe posting that or sending it along to Orion support would be helpful. Weblogic certainly are very keen to get thread dumps with these kind of issues, and there's no reason to think that it'd be any less useful in the case of Orion!
Finally, if you can identify an exact sequence that would cause the crash, you're halfway to solving it. You've mentioned that logging out of the machine would crash the JVM, that kinda points the finger at the remote control software you're using.
Also a query that returns 10,000 hits is not so bad, and pretty much any JVM should be able to deal with that without much trouble. Just make sure you're allocating plenty of heap memory for it. (using -mx 256M -ms 256M)
> -----Original Message-----
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED]]On Behalf Of
> Duffey, Kevin
> Sent: Thursday, October 05, 2000 2:11 PM
> To: Orion-Interest
> Subject: RE: WinNT 4, SP5/6 and Orion consistently crashing..
>
>
> Sorry..I was a bit vague. What appears to happen is that
> Orion completely
> shuts down. I am not sure if you are anyone is familiar with
> remote access
> to administer a computer. What I do from my local box, using either
> PCAnywhere (for local computers) or Remote Desktop (for our production
> systems because the co-lo facility forces us to use that
> software). I log in
> to the system. A few things happen. First, if we just log out of the
> computer normally, it seems to crash any Java running
> app..actually..it
> causes a thread dump (or core dump..not sure) of the JVM. We
> often have to
> log back in and restart. This we noticed with IIS/JRUN. In
> hopes that SP6
> would solve the problem, we upgraded to SP6 on one machine.
> Again..this is
> NT4 with SP6 and JDK1.2.2. Now, I am using run_as_service to
> run Orion as a
> service. It seems to work fine..so I am not sure if that
> program causes a
> problem. I wish there was some OS level way to run Java as a
> service without
> having to use a 3rd party tool. Sucks that they don't support
> that. I wonder
> if Win2K does that (do you know?). Anyways...it appears
> though, that when
> another person logs in, then I come back in a bit later, the
> Orion window is
> gone. The service says its still running, so I have to STOP
> the service then
> START it again. But nobody can connect to the site any
> longer. I start the
> Orion service with java.exe not javaw.exe so I can see the
> output. We do
> some output of connection check-out/check-in, etc. Since we
> aren't deployed
> on Orion yet, we haven't even tried to convert to using
> Orion's connection
> pooling. I will say this much, so as not to point blame at
> Orion because I
> really don't think that is the problem. Our code has a home-brewn
> persistence solution, home-brewn faxing solution (we are
> basically a company
> that does about 3500 faxes a day..we have a pool of fax
> modems, and a faxing
> program called HylaFax which we are replacing with a new one
> called Replix
> and using an out-sourced faxing farm). I don't know a lot
> about JDBC myself,
> so I am not sure how we are storing objects, retrieving them
> etc. I have
> worked with our queries a bit, the "construction" of the WHERE clause
> anyways. I am also skeptical of why a machine with 512MB of
> ram runs out of
> memory on a single query. The query can return results of
> 10,000 station
> objects, each object can't take up more than 1 or 2K of
> memory. Even worse,
> on a E450 server (our Oracle runs on that) with 4 400Mhz cpus
> (64-bit), 4GB
> memory and a Veritos cluster of very fast hd's, it takes 7
> minutes to get
> 464 station objects back in a query. From the query it
> doesn't appear that
> we are doing many joins. All of this leads me to believe our
> code is written
> poorly (thankfully I wasn't involved in that code), and we
> have a lot of
> work to do to figure out why we run out of memory, queries
> take a long time,
> etc. But, I still haven't a clue why the Orion server would randomly
> disappear from the screen and we suddenly can't access it
> anymore. Another
> thing we notice is that when a person submits a query that
> seems to take a
> long time, Orion drastically degrades in performance. My
> thinking is that
> one thread running the query takes over, and other threads of the same
> priority (each incoming request) don't get executed very
> quickly. The CPU
> utilization goes to 100% when that happens, so I don't know
> if its the query
> just went bad, or what. We are using JDBC 1.0 by the way.
>
> So..I am sort of up in arms about what to do. The worse part
> is we are not
> given the time to diagnose these issues. We keep moving forward with
> business needs and soon we expect to have 3 to 4 times the
> traffic on our
> site, and I don't think we are prepared to handle it. So, I
> guess I should
> say if you can shed any light on some areas of trying to
> diagnose these
> issues..that would be great. One thing I am doing..I am
> running Orion as an
> application on a Win2K box with JDK 1.3. I am going to see if
> it suffers at
> all from the same symptons we were seeing at our co-lo with
> NT4/sp6. If not,
> then that would give us the answer..at least to why it was
> crashing. If it
> does crash, then either Orion, our code or both are at fault,
> and we'll have
> to get to the bottom of that.
>
> Thanks Karl, Magnus and anyone else with feedback.
>
>
>
> > Hello Kevin,
> >
> > What do you mean by crashing? Does Orion exit, do you get a
> > Dr. Watson, do you
> > get any other error, does it just stop listening to requests?
> >
> > Regards,
> > Karl Avedal
> >
> > "Duffey, Kevin" wrote:
> >
> > > Hi all,
> > >
> > > I am not sure why..but it appears that Orion 1.2.9 (and
> > 1.3.7) crashes
> > > whenever we "remote" into our server to do some work on it.
> > We have NT4 with
> > > SP5 on one machine, and SP6 on another. I have Orion using
> > run_as_service to
> > > put it in as a service. Even when I run it as an
> > application, this happens.
> > > It seems to work fine for a while, then it crashes. I know
> > our code is bad,
> > > but it seems to me with only 4 people using the thing for
> > testing, it should
> > > be fine. Has anyone been experiencing problems with Orion
> > crashing on NT
> > > with SP5 or SP6? Does the remote access do something to the
> > JVM? Its using
> > > JDK 1.2.2-05 JRE.
> > >
> > > On the subject..does anyone know of a Win2K program that
> > will put Java apps
> > > as services? The run_as_service doesn't appear to
> > work..keeps giving some
> > > i/o problem.
> > >
> > > Thanks.
> >
>
