I've had no problem with IBM JDK 1.3.0 on redhat 6.1 with glibc 2.1.3
kernel 2.2.12 I posted some tests a while back with multi threaded
(multi machine) load testing, I had some problems with tomcat 4
at the time but I 3.2 was fine.
I'm currently setting up a redhat 7 machine glibc 2.2 IBM 1.3.0 I
haven't run much load testing on it so far, have you tried the
newler glibc on your redhat 7 machine?
Bill Graham wrote:
>
> I've been doing a lot of similar tests in search of the best jdk to
> work with linux and tomcat and have been seeing the same lack of
> stability as Chris. I was running the IBM JDK 1.3.0 bld cx130-20010207
> with glibc-2.1.3-15 and kernel 2.2.16-21.7.1 running with both SMP
> and UP. The VM would just hang after a few minutes of heavy load.
>
> It seemed similar to this bug mentioned in the IBM.Tomcat.Linux group
> list (although their bug was seen with jvmpi):
> http://groups.google.com/groups?hl=en&lr=&group=ibm.software.java.linux&safe
> =off&ic=1&th=443066d336bfa758&seekd=909050309#909050309
>
> Sun's VM wasn't any better:
> This hanging thread bug reported against Sun's 1.2.2 VM is still open and
> looks like the same thing that I've seeing with the 1.3.0 VM:
> http://developer.java.sun.com/developer/bugParade/bugs/4306523.html
>
> I tried running the Sun's VM in in classic mode (green threads) and was
> easily able to reproduce their SEGV bug:
> http://developer.java.sun.com/developer/bugParade/bugs/4372197.html
>
> We've searched extensively for a good vm for linux and haven't found one
> that can pass the test. (Our Solaris machines pass with flying colors,
> albeit a bit slower.)
>
> I've heard some people on this list say that they've had success with
> linux, but none that have done any multithreaded load testing. If anyone
> does have a winning combination, please pipe in!
>
> bill
>
> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of David Crooke
> Sent: Saturday, April 07, 2001 9:58 AM
> To: [EMAIL PROTECTED]
> Subject: Re: Tomcat3.2+Redhat 7.0+glibc-2.x.x+jdk1.x == stable??
>
> The glibc that RH7 shipped with was beta and very broken - you need to apply
> their patch set. Red Hat rushed their 7.0 release, breaking their
> convention (major version = new Linux kernel) and a lot of stuff besides.
> Most of the new stuff in it is GUI related, and as such you may be
> better sticking with 6.2 for server use.
>
> Depending on what your code does, you may get better performance from a
> green threads VM - with 1.3 these are starting to go away; you can still
> get green threads with the Sun JDK using --classic but I believe it hobbles
> HotSpot. We still run 1.2.2 in production and are evaluating the 1.3
> VM's for stability.
>
> Like most open source projects, Apache stuff comes from a Unix heritage, and
> while you can shoehorn it onto Windows, you will have a much easier
> life if you stick with Linux. Just check this newsgroup - 90% of the traffic
> is "I can't get x, y or z to work on NT"
>
> Chris Price wrote:
>
> > Hi;
> >
> > I know its kind of an open ended question, but I would like some
> input
> > (factual or anecdotal) on other peoples success in running various
> > combinations of glibc + jdk + kernel (does kernel mattrer much?).
> >
> > I am currently running the Sun JDK 1.3.0 with glibc-2.1.92 and
> kernel
> > 2.2.16-22 on a *very* stock RH 7.0. This combination is proving itself
> > to be quite unstable, with the Tomcat http listener and/or servlet
> > engine dying under moderate loads (less than 4,000 hits roughly over 8
> > hours).
> >
> > The glibc + kernel are stock RH7.0 pieces, and while I am sure I
> can
> > probably gain some stability by upgrading to newer ones, I would like to
> > know if there is a "optimal" (used loosely) combination that exists.
> >
> > I am inheriting this setup from an Admin that thought that Linux
> > 'sucked' and that we should move to NT. The previous admin didn't know
> > how to use rpm, compile a kernel, or have a clue as to what glibc is.
> > Needless to say, I can probably add some value and squeeze some better
> > performance from our Tomcat app servers.
> >
> > Cheers
> >
> > Chris Price
