We would have made exactly the same observation as Jeff and his team. As my
last post indicated, on further review (moving windows around for better
feedback, and doing things more slowly) it becomes clear that the
com.evermind.server.http.HttpIOException happens first, and with both the
ATM and News demos is completely reproducible when running on the 1.2.1_04
Solaris JVM.

_If_ we do certain things _after_ seeing the exception, then various signals
may be raised - SIGBUS usually. But the exception invariably happens.

We are taking your advice to heart, and have downloaded the 1.2.2_05 Solaris
JVM and patches, but I get the impression from what Jeff says that this may
not help.

I guess the real question that both of us have is this: in the complete
absence of a SIGBUS or SIGSEGV or anything else, why would
"com.evermind.server.http.HttpIOException" indicate anything else other than
what it purports to be; an exception thrown by Orion? I don't see that this
exception indicates anything amiss with the JVM at all, in and of itself.

Regards, Arved Sandstrom

-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED]]On Behalf Of Bailey, Jeff A
Sent: Monday, May 08, 2000 3:35 PM
To: Orion-Interest
Subject: RE: EJBs Problems with Orion on Solaris 2.7


Karl,

Thanks for the quick response.

We assumed (by we I am referring to my development staff) that the SIGBUS
10* bus error was definately JVM related, and determined that things such as
this are not the result of Orion, but you did not address the
com.evermind.server.http.HttpIOException: Read timed out errors that are
occuring 10 to 15 seconds after each page request.  We have performed a
number of JVM updates with no luck in removing this Exception.

We are not getting the SIGBUS error at all and the only error we are getting
is a log full of stack traces similar to what was originally posted (which
tends to grow fairly large over time).

The error does not seem to have much affect on Orion's ability to serve
pages at a low to moderate traffic level, but I do now know what impact this
error will have on high traffic levels.

Thanks

-j

> -----Original Message-----
> From: Karl Avedal [mailto:[EMAIL PROTECTED]]
> Sent: Monday, May 08, 2000 11:17 AM
> To: Orion-Interest
> Cc: Orion-Interest
> Subject: Re: EJBs Problems with Orion on Solaris 2.7
>
>
> Hello,
>
> I guess you all already know that a SIGBUS 10* bus error is
> not a Java error,
> but a native error, which means that it is caused by a JVM
> bug that is pretty
> much out of our control. We try to work around JVM bugs as
> much as possible,
> but obviously, some JVM bugs are fatal. When you get hit by a
> JVM bug, this is
> what you should do:
>
> 1. Get the absolutely latest version of your JVM.
>
> 2. If the latest version is still buggy, report the bug to
> your JVM vendor and
> send a mail to [EMAIL PROTECTED] We will try to work
> around the bug if we
> can reproduce it, but there are no guarantees that this is
> doable. If it is not
> doable we can at least keep track of which JVMs are buggy
> which will help us
> help you. After reporting the bug to the VM vendor, try a JVM
> from another
> vendor. If this fails too, repeat the steps again.
>
> The Solaris JVMs are unfortunately known to contain many bugs
> that affect
> server software. Do try the latest release to see if this
> fixes your problem.
>
> Good luck!
>
> Regards,
> Karl Avedal
>
> PS. How do you know if it's a JVM bug? Well, if what you get
> does not look like
> an Orion error message or a stacktrace it is probably a JVM
> bug. If the error
> causes the Orion process to die, it's almost certainly a JVM
> bug. If you get
> core dumps, segmentation faults, bus errors, or other OS
> error messages, it's
> definately a JVM bug (or possibly an OS bug, but that's much
> less common).
> Orion does not contain native code which means that it can
> not create native
> errors.
>
> Arved Sandstrom wrote:
>
> > I reran the ATM more carefully. What happens is actually
> what you describe;
> > the "com.evermind.server.http.HttpIOException: Read timed
> out" occurs no
> > matter what. I estimate maybe 10 seconds after the page is
> requested, so
> > it's similar to what you see.
> >
> > The SIGBUS happens after the above when trying to do
> another page request.
> > So it's not the problem - it's a result of the problem.
> >
> > I'm baffled. :-)
> >
> > Arved
> >
> > -----Original Message-----
> > From: [EMAIL PROTECTED]
> > [mailto:[EMAIL PROTECTED]]On Behalf Of
> Bailey, Jeff A
> > Sent: Monday, May 08, 2000 12:20 PM
> > To: Orion-Interest
> > Subject: RE: EJBs Problems with Orion on Solaris 2.7
> >
> > We get the same error on Solaris 2.7 with the exception
> that we do not get
> > the SIGBUS 10* bus error.  Instead, about 12 to 15 seconds each page
> > request, we get the exception you have listed below.  This
> doesnt occur on
> > any other OS, including Intel Linux, Win NT, and Win 2000.
> >
> > Hopefully this helps,
> >
> > -j
> >
> > > -----Original Message-----
> > > From: Arved Sandstrom [mailto:[EMAIL PROTECTED]]
> > > Sent: Monday, May 08, 2000 7:58 AM
> > > To: Orion-Interest
> > > Subject: EJBs Problems with Orion on Solaris 2.7
> > >
> > >
> > > I've used Orion with success on Windows 98, NT and Intel
> > > Linux. I'd like to
> > > also use it on Solaris 2.7.
> > >
> > > I set up an Orion installation with 1.0rc1 in my HOME
> > > directory. The JVM in
> > > use is:
> > >
> > > $ java -version
> > > java version "1.2.1"
> > > Solaris VM (build Solaris_JDK_1.2.1_04c, native threads, sunwjit)
> > >
> > > I set a <library path="..." /> to the tools.jar for that JVM.
> > >
> > > Since I'm not superuser I set the HTTP port to 8080. I'm
> > > using the latest
> > > version of HypersonicSQL.
> > >
> > > The default examples (JSPs) work fine. The ATM and News apps
> > > deploy OK, and
> > > I can do things with both up to a certain point. At a certain
> > > point with
> > > both (after creating a new account with the ATM app, and
> > > trying to add a new
> > > news item with News) Orion throws the following:
> > >
> > > com.evermind.server.http.HttpIOException: Read timed out
> > >         at java.lang.Throwable.fillInStackTrace(Native Method)
> > >         at java.lang.Throwable.fillInStackTrace(Compiled Code)
> > >         at java.lang.Throwable.<init>(Compiled Code)
> > >         at java.lang.Exception.<init>(Compiled Code)
> > >         at java.io.IOException.<init>(Compiled Code)
> > >         at com.evermind.server.http.HttpIOException.<init>(JAX)
> > >         at
> > > com.evermind.server.http.EvermindHttpServletRequest.cp(Compiled
> > > Code)
> > >         at com.evermind.server.http.dl.cd(Compiled Code)
> > >         at com.evermind.util.e.run(Compiled Code)
> > >
> > > This is accompanied with a SIGBUS 10* bus error.
> > >
> > > Has anyone run into this kind of stuff? I don't have anything,
> > > configuration-wise, set up any different, in principle, than
> > > what has worked
> > > successfully on all 3 of the previously-mentioned platforms.
> > >
> > > Thanks in advance.
> > >
> > > Arved Sandstrom
> > >
> > >
>
>



Reply via email to