Is there any way to stop that module from loading without having to
modify the initrd?
dont know that, but actually just for sake of experiment it is not hard
to make new ramdisk archive, for example like that for the running kernel
bash# mkinitramfs -o /boot/ramdisk-20080705.img
and direct additional grub boot entry to use it.
In this way you could at least check how different drivers perform with
your adapter. To look into ramdisk just unpack the archive
bash# mkdir /tmp/ramdisk-20080705 && cd /tmp/ramdisk-20080705
bash# gunzip -c ../ramdisk.img | cpio -dmvi
Meanwhile, i've modified my xend-config file, setting dom0-cpus to 1 (it
was set to 0). Then i restarted the server, and so far so good. I'll
keep it running for several days and see what happens.
The network adapter always works fine for some time when the system has
been restarted. And then, all of a sudden, it starts performing very
poorly, or it looses connection, having to reboot in order to get it
Somehow i think there is some kind of unluck going on around your
motherboard and adapter and perhaps easiest workaround is just to get
new ethernet adapter and life is good again! :)
Or as an experiment boot again your system up with an ordinary kernel
and stresstest network adapter. For example flood some traffic thru it
with hping3 and monitor how it performs with iptraf.
And last but not least, Xen system shows up for the local switch as
multiple mac addresses behind the same switch port. Maybe it has
something to do with it. If you can, use only single ip address and mac
address in association with that adapter. I.e. in dom0 leave the ethx's
ip addres unconfigured and configure its ip address only in domU.
Also, perhaps you could load ethernet adpater module with some arguments
or tweak something with ifconfig i.e. media options. With ethtool -s you
could also change media options like speed, maybe it changes it somehow
And in ifconfig output look at the errors
RX packets:132854764 errors:0 dropped:0 overruns:0 frame:0
TX packets:133486462 errors:0 dropped:0 overruns:0 carrier:0
I am quite ordinary computer user, thats about all what i could possible
help on this one, some network or xen hard core is of more help most
probably. And good luck!
PS I have old good asus a8n-sli premium motherboard which has lost by
now almost everything but its name, i.e. onboard sata ports start giving
errors after running some weeks and also onboard ethernet adapters stop
working. At the same time ethernet adapters as pci card work correctly
although sometime they agree with switch on wrong speed.
2008/7/5 Imre Oolberg <imre@xxxxxxxxxxx <mailto:imre@xxxxxxxxxxx>>:
If it helps but i found out which driver is associated with my
p/ethx devices running ethtool like this
aix:~# ethtool -i peth1
And to make sure wrong driver isnt getting in the way i sometimes
just removed it from module tree under /lib/modules/... (or ramdisk
i.e. initrd image) though there is obviously more appropriate way
I also run the same software and only thing which made it unstable
was having multiple cpu's in dom0 but it is know fact and workaround
is to have in /etc/xen/xend-config.sxp
# In SMP system, dom0 will use dom0-cpus # of CPUS
# If dom0-cpus = 0, dom0 will take all cpus available
Alejandro Martini wrote:
Randall, thank you for sharing those tips. The network card was
working fine on this system before installing Xen. At least i've
never noticed any problems like the ones i reported while i was
setting up the base system (Debian Etch, via netinstall).
Now i think that the problem might be related to the fact that
Xen is not loading the right module for this network adapter.
I've googled around quite a bit, but wasn't able to find a
conclusive statement on what kind of card this is.
There are at least three possibilities, since:
- The manufacturer states that this is a Realtek RTL8111B (the
motherboard is an ASRock AliveNF5-eSATAII)
- lspci identifies it as "Realtek Semiconductor Co., Ltd.
RTL8111/8168B PCI Express Gigabit Ethernet controller"
- udev seems to have detected it as 8169
- Xen loads the r8169 module for this card, even if r8168.ko is
present and r8169 is blacklisted.
So, i cannot tell for sure if i'm dealing with an 8169, an 8168B
or an 8111B.
If i run modinfo, i can see that both modules (r8168 and r8169)
have matching aliases for this card type (10ec:8168), and then i
don't really know how the kernel decides which module to load,
or how to tell it to choose one over the other.
With regard to your other advice (installing irqbalance), the
latest known version for debian/stable is 0.12-7. I guess that
it's pretty outdated. Anyway, if you think that it's worth
trying, i'll install it and see what happens.
2008/6/23 randall ehren <randall@xxxxxxxxxxxxx
> two things i can think of:
> 1) did you ever test this machine's networking performance
before installing xen?
> 2) have you tried installed 'irqbalance'? i've had to
install that on dell 2650's running xen 3.0.x
Xen-users mailing list
Xen-users mailing list
Xen-users mailing list