On Mon, Oct 05, 2009 at 05:25:15PM -0400, Chris wrote:
> Quoting Joachim Schipper <[email protected]>:
> >On Mon, Oct 05, 2009 at 01:16:07PM -0400, Chris wrote:
> >>I just reinstalled 4.5. I touched nothing [but] I installed mutt
> >>through pkg_add, then created a 1M empty file from /dev/null.
> >>
> >>I sent this email to myself thusly: mutt -a 1megfile [email protected]
> >></dev/null
> >>
> >>(...) It spools, then I have about .5 - 1 second
> >>before then entire system locks up.
> >
> >>(...) I am running this on a Xen virutal server (on Debian Lenny)...
> >The OpenBSD stance on virtualization is pretty much "don't" (...)
> >[But] it is possible.
> >
> >(...) OpenBSD can only run as a guest with hardware
> >support (Intel VT/Vanderpool or AMD-V/Pacifica); so-called
> >paravirtualization doesn't work. If you do not have the required
> >hardware - and you can see whether or not you have it in the Xen logs -
> >you could use qemu[, which is] known to work.
> >
> >I'm not sure if Xen uses the hardware support by default; you might want
> >to look into that, too.
> >
> >[T]ry compiling /usr/src [to] test of a large number of subsystems.
> Well, it ran poorly under vmware... and it seemed to work great under
> Xen, until this showed up. Everything else worked fine -- it is just
> the sendmail issue, and only with large emails. I can move large
> files back and forth just fine via http and sftp.
>
> Yes, I do run it as an hvm, and I have an AMD/Pacifica compliant chip
> -- no issues there. It is rock solid with this one exception. I have
> another built as web server, and again, no issues.
>
> I'm not entirely sure what you meant by "hardware support", when you
> say "not sure if xen uses the hardware support by default".
>
> My biggest problem is that I cannot get it to create any error output
> anywhere, not in the xen logs, not in the qemu logs, not in
> /var/log/messages (on either dom0 or domu), not with sendmail running
> in the foreground and not in /var/log/maillog (on domu)... It just
> dies Very frustrating!
>
> I was following the stable branch, and I have actually compiled
> /usr/src -- twice now, it was about 18-20hrs of compiling but went
> without a hitch. It is just this one issue that takes it down - 100%
> of the time.
>
> Maybe if you can clarify what you meant by "hardware support", I can
> follow that lead... otherwise, I... I think I'm hosed...
>
> I appreciate the input. Thanks.
Hardware support, indeed, means something like Pacifica. So apparently
that isn't the issue.
It's rather disappointing that you don't see any log output at all; that
makes it rather hard to debug the issue. Still, it's very surprising
that sendmail and only sendmail causes this crash; sendmail isn't *that*
special.
Just to clarify: have you tried running httpd or sshd on the guest
system? httpd, in particular, does the same "take things from disk and
stuff them onto the network" as sendmail. (I fail to see why running the
server end on the Xen guest instead of host would make a difference,
but...)
I hope someone with more knowledge weighs in; I must admit that I don't
know what could be wrong. Maybe a Xen list could be of more assistance?
Joachim
