Xen devel - Feb 2011 - Abnormal memory usage of dom0 under Squeeze leading to crash

Hi,

We normally run Lenny dom0 with 384 MB of RAM in our production server.
But as we did few trials with Squeeze, I had the very bad finding that a
Squeeze dom0 requires at least 1024 MB of RAM. As soon as I do something
like "xm mem-set Domain-0 800", my server crashes: OOM starts killing
processes, starting with sshd, then mdadm and so on, then the server
reboots by itself. This enormous amount of RAM isn''t normal at all, and
I am wondering what is happening.

If I sort processes by memory usage using top, what I can see is 2 xend
processes taking 1.0 and 1.5% of the RAM (out of 1024 MB). This doesn''t
seem to be so much, and I don''t understand. My server really
shouldn''t
behave like this! (we even used to run dom0 with 192 MB of RAM back when
we were using Xen 2.0.7 and sarge...) My concern isn''t even the waste
of
RAM, but a danger for the stability of the system. It just doesn''t feel
right!

I have attached the output of dpkg -l, and my setup script is available
here:

wget
"http://git.gplhost.com/gitweb/?p=mgmt-scripts.git;a=blob_plain;f=scripts/dom0-postinstall-squeeze;hb=2c869ce9d729d80a687f1fbdbf0081e140152b44"
-O dom0-postinstall-squeeze

With both of the above, it''s going to be easy to reproduce. As Squeeze
is said to be out in 3 days, it would be great to find out what''s going
on and fix asap.

Does anyone have an idea of what''s going on? I just don''t get
it...

Thomas Goirand (zigo)

[ Stripped debian-release from the recipients, this list is not for
dicussion ]

On Thu, Feb 03, 2011 at 01:37:16PM +0800, Thomas Goirand
wrote:
> We normally run Lenny dom0 with 384 MB of RAM in our production server.
So do I:
| $ sudo xm dmesg | grep Comm
| (XEN) Command line: dom0_mem=384M
| $ sudo xm info
| [...]
| release                : 2.6.32-5-xen-amd64
| [...]
| xen_major              : 4
| xen_minor              : 0
| xen_extra              : .1
| [...]
> But as we did few trials with Squeeze, I had the very bad finding that a
> Squeeze dom0 requires at least 1024 MB of RAM. As soon as I do something
> like "xm mem-set Domain-0 800", my server crashes: OOM starts
killing
> processes, starting with sshd, then mdadm and so on, then the server
> reboots by itself. This enormous amount of RAM isn't normal at all, and
> I am wondering what is happening.
Way not enough information. You are actually balooning the dom0 down?

Please provide all informations. xm dmesg, the kernel log, xm info.
> With both of the above, it's going to be easy to reproduce.
Nope. Works fine here.
>                                                             As Squeeze
> is said to be out in 3 days, it would be great to find out what's going
> on and fix asap.
There will be no fix for the initial release.
> Desired=Unknown/Install/Remove/Purge/Hold
> |
Status=Not/Inst/Conf-files/Unpacked/halF-conf/Half-inst/trig-aWait/Trig-pend
> |/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad)
> ||/ Name                                    Version                    
Description
[...]

This looks more like a workstation then a minimal vm host.

Bastian

-- 
Star Trek Lives!