Xen users - Sep 2005 - Kernel panic recoverable by hitting <enter>?

Hi *,

I''m running several xenified systems [UP|SMP]/[P4,Athlon,Xeon] from 
multiple maufacturers [HP,Intel,IBM,Dell] in different CoLos since early 
2004 24/7 in test as well as production scenarios without any noteworthy 
problems.

But, suddenly, a Dell PowerEdge 1750 equipped with SMP Xeon@2.80GHz and 
4GB DDR ECC RAM (from crucial.com) running under xen-2.0.5 with 
2.6.10-xen0 on Debian Sarge and carrying 22 domUs [Sarge|FC4|*BSD] under 
medium load disappeared this afternoon from the network - both the dom0 
and all domUs.

The remote-hands service claimed that there was a kernel panic on the 
system''s console showing loads of hexdumps and something related to 
''memory stack'' - unfotunately, the guy couldn''t
remember exactly what
was shown on the screen [and, sadly, I''ve no serial console attached to
that machine and no terminal server, of course ;-)].

Now the odd part, at least for me as I never heard of such a thing: the 
technician also claimed that, after pressing the enter key, the system 
returned to ''normal operation'' and showed a conosle login.

When the machine reappeared back on the net it looked like as if nothing 
had happened: Uptime of dom0 is still >90 days, neither the logs of dom0 
nor those of the domUs show any suspicious entries - except some of the 
domU''s services complaining about the lack of network connectivity 
(openvpn, etc...).

Do you guys have heard of user-confirmable kernel panics? (I just 
grepped through the xen- and linux-sources but found nothing; Google 
wasn''t of much help, either.)

Was the remote-hands-guy telling me fary tales merely to draw off 
attention from some colo-internal network problems? Or could that be a 
xen-specific experience? How to track that down?

Regards,

Tobias

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

> But, suddenly, a Dell PowerEdge 1750 equipped with SMP Xeon@2.80GHz and
> 4GB DDR ECC RAM (from crucial.com) running under xen-2.0.5 with
> 2.6.10-xen0 on Debian Sarge and carrying 22 domUs [Sarge|FC4|*BSD] under
> medium load disappeared this afternoon from the network - both the dom0
> and all domUs.
Nice setup!
> The remote-hands service claimed that there was a kernel panic on the
> system''s console showing loads of hexdumps and something related
to
> ''memory stack'' - unfotunately, the guy couldn''t
remember exactly what
> was shown on the screen [and, sadly, I''ve no serial console
attached to
> that machine and no terminal server, of course ;-)].
Shame they didn''t just take a digicam photo and mail it to you.
> Now the odd part, at least for me as I never heard of such a thing: the
> technician also claimed that, after pressing the enter key, the system
> returned to ''normal operation'' and showed a conosle
login.
>
> When the machine reappeared back on the net it looked like as if nothing
> had happened: Uptime of dom0 is still >90 days, neither the logs of dom0
> nor those of the domUs show any suspicious entries - except some of the
> domU''s services complaining about the lack of network connectivity
> (openvpn, etc...).
But all the domUs will still up?
> Do you guys have heard of user-confirmable kernel panics? (I just
> grepped through the xen- and linux-sources but found nothing; Google
> wasn''t of much help, either.)
They don''t exist...  But that assumes it was a panic.  The message he
saw
could have been some sort of non-fatal oops - have a look in dom0''s
dmesg
output.  I guess it might have somehow recovered from <whatever it was>
and
thus returned your connectivity.  Don''t ask me what it could be though
;-)
> Was the remote-hands-guy telling me fary tales merely to draw off
> attention from some colo-internal network problems? Or could that be a
> xen-specific experience? How to track that down?
As well as checking the dmesg in dom0, you might also like to check for 
suspicious warnings in xm dmesg - just in case Xen spotted a domain doing 
something weird and complained about it.

Other than that, it seems a bit mysterious...

Cheers,
Mark

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users