[xenpreview-dev] Stability & Bugs

[Jan Albrecht]

Hi,

it seems to me that the stability of XEN (full virtualization) with the
latest changeset is much better than with the original packages of
SLES10, but not as good as I thought.

I've this enviroment:
HP DL380G5
1 Dual Core CPU model name      : Intel(R) Xeon(R) CPU            5120 
@ 1.86GHz
3 GB RAM
SAS Disks

I've some Debian guests, one Windows 2003 Server and one SLES 10 guest.
All are full virtualized.

When I do some heavy memory tests (let's call the malloc), where just
the memory is allocated every two seconds (2 processes with 64 MB) and
freed again, the XEN host server dies unexpected. It just stops working,
I can try to enter "root" at the console and then it just stops.
Rebooting does work.
(If anyone is interested in my malloc script I will provide this.)

Or an another example: I do restart the SLES10 instance via
"ctrl-alt-del" with "F8" in the VNCviewer window. The restart does work
fine until the system does reach the udev. Then it does slow down heavy
and the CPU usage in "xm top" does go up to nearly 100%. Services such
network do not work when this does happen.
A system halt of the guest does resolve this problem.
(I can attach a screenshot, but this will exceed my mailsize limit in this list...)

Also starting a second instance of an already running instance is
possible. Earlier versions of "xm create" did prevent that, the current
changeset does not.

Any ideas about this?

Jan