Xen users - Mar 2012 - "Hard" hangs on several Intel(-chipset) boards

Hey!

I''m deploying a vanilla Xen 4.1.2 installation at a hosting provider, 
and I''m experiencing hard hangs (i.e., the system is unresponsive to 
keystrokes, network is off, I/O to disk is stopped) on several of their 
available server packages, all based on Intel i7 CPUs and related chipsets.

The failure does not seem related to a specific Xen-version (I''ve 
experienced the same hangs with all [vanilla] Xens from 4.0.1 to 4.1.2), 
not related to the kernel (I''ve experienced the same hangs with a 
Xen-ified 2.6.38 from SuSE, and several vanilla 3.2-kernels with 
pv_ops), and not to the actual hardware in use (I''ve seen the same kind
of hangs on an MSI X58 Pro-E [with an i7-950] and after swapping 
hardware also on an ASUS P8H67-M PRO [with an i7-2600]). The additional 
hardware specs (RAM type, Disks) are also different between several test 
systems (i.e., I''ve had access to systems with Samsung, Kingston, and 
F3-RAM).

As I''ve not been able to get a serial cable attached for in-depth 
debugging of Xen problems on the hangs (the hosting-provider is not very 
cooperative), is there anything else I can do to debug these hangs? Or, 
has anybody else experienced the same hangs?

Thanks for any hint!

-- 
--- Heiko.

Hi Heiko,

On Mon, Mar 12, Heiko Wundram wrote:
> As I''ve not been able to get a serial cable attached for in-depth
> debugging of Xen problems on the hangs (the hosting-provider is not
> very cooperative), is there anything else I can do to debug these
> hangs? Or, has anybody else experienced the same hangs?
Which distribution do you use for dom0 ?
I had the same hangs (after about 1 day) with opensuse 12.1 dom0.
With opensuse 11.4 as dom0 I didn''t had any hangs.


-- 
Best regards

  Dieter

--
I do not get viruses because I do not use MS software.
If you use Outlook then please do not put my email address in your
address-book so that WHEN you get a virus it won''t use my address in
the
From field.

Am 12.03.2012 12:47, schrieb Dieter Bloms:
> Which distribution do you use for dom0 ?
> I had the same hangs (after about 1 day) with opensuse 12.1 dom0.
> With opensuse 11.4 as dom0 I didn''t had any hangs.
The Dom0 is a (relatively) current Gentoo ~amd64; the 2.6.38-style 
kernel is a rebased OpenSuSE 11.4 (IIRC) xenified kernel, which comes as 
part of the xen-sources package of Gentoo. Xen (and related packages) 
come from Gentoo Portage (no "real" source compilations), which are 
vanilla besides a few patches for autoconf/automake stuff.

-- 
--- Heiko.

Hello.

El 12/03/12 02:28, Heiko Wundram escribió:
> As I''ve not been able to get a serial cable attached for in-depth
> debugging of Xen problems on the hangs (the hosting-provider is not very
> cooperative), is there anything else I can do to debug these hangs? Or,
> has anybody else experienced the same hangs?
For debugging, if there is no IPMI or IP-KVM and syslog to remote server 
does not help, you may try to get a "netconsole".

This is kinda rustic, the sender is set up via a kernel boot argument, 
the receiver is a raw network socket (netcat works). No authentication, 
no encryption, no routing (sic! must be within same IP subnet).

Check this links:
http://cateee.net/lkddb/web-lkddb/NETCONSOLE.html
http://lxr.linux.no/#linux+v3.2.9/Documentation/networking/netconsole.txt

Greetings.

-- 
Alexandre Kouznetsov

Am 12.03.2012 17:40, schrieb Alexandre Kouznetsov:
> This is kinda rustic, the sender is set up via a kernel boot argument,
> the receiver is a raw network socket (netcat works). No authentication,
> no encryption, no routing (sic! must be within same IP subnet).
>
> Check this links:
> http://cateee.net/lkddb/web-lkddb/NETCONSOLE.html
> http://lxr.linux.no/#linux+v3.2.9/Documentation/networking/netconsole.txt
Thanks, that''s a wonderful hint! I have an IPMI-console attached to the
system so getting screen output is not my current problem, but the 
screen doesn''t show any backtrace from the Dom0 (and I''ve
already turned
consoleblanking off), so possibly I''ll get some output using the
netconsole.

-- 
--- Heiko.