FreeBSD 5.2.1 kernel w/ SMP under high load = panic

[Chuck Rock]

I had a similar problem with this release running on any Dell 2550 Rack
server I had. I tried three different servers with the same config, and I
had kernel panics at various times. All of them had different information
on the screen at each panic.

I had several people tel lme they had no such problems with the same
systems.

I chalked it up to bad driver for the PERC3/DC raid controller. All panics
seemd to happen under very high disk I/O. The machine would run fine for a
few days, then just puke.

I moved all the stuff over to a refurb Compaq Dual Xeon 2.8, and have had
only one problem lockup in the 5 months I've been running it under
increasing loads. The last lockup I had also seemed related to disk I/O.

Otherwise, 5.2.1 has been a very stable production server processing over
1 million E-mail's a day incoming and hundred thousand out.

Chuck

On Mon, 2 Aug 2004, John Baldwin wrote:

> On Sunday 01 August 2004 09:08 am, Hugo Silva wrote:
> > I am running a Dual Xeon 2.8ghz w/ SMP, SCHED_ULE on FreeBSD
> > 5.2.1-RELEASE-p9.
> >
> > Whenever I load the server a bit more (let seti at home run, compile stuff in
> > multiple jails, etc), it will simple go offline. I confirmed with the
> > datacenter and it is indeed a panic, but the datacenter didn't give me the
> > panic message. I know for *sure* it's because of the high loads.
>
> Hmm, we'd really need the panic message to even start debugging it.
>
> > I need to sort this out, this is a powerful server being cut because of
> > FreeBSD/SMP, and I know there is a kernel option to prevent the panic, I
> > read about it ages ago on a forum. But I can't locate it. That user said
> > if he disabled SMP, panics would stop. But another user suggested adding a
> > kernel option (which I simply don't remember), and panics stopped, even
> > with SMP.
> >
> > I tried KVA_PAGES=512, but it only caused another panic, this time as soon
> > as the system started up..
> >
> > syncing disks, buffers remaining... panic: pmap_invalidate_range:
> > interrupts disabled
> > cpuid = 0;
> > boot() called on cpu#0
> > uptime: 9s
>
> Unfortunately, I'd really need the backtrace to see how to fix this panic.
>
> --
> John Baldwin <jhb at FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
> "Power Users Use the Power to Serve"  =  http://www.FreeBSD.org
> _______________________________________________
> freebsd-smp at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-smp
> To unsubscribe, send any mail to "freebsd-smp-unsubscribe at freebsd.org"
>