On Sat, Jan 29, 2011 at 07:03:16PM +0100, Jordan Pittier wrote:
> Thanks for your reply. LSI has indeed newer driver for the controler;
> but I can't "build" it, there's an error when I try to compile it [see
> attachement]. I will give another try in the next days.
> What is puzzling is that the IO errors only occurs with Xen HV. I am
> 100% willing to accept that the problem is the drivers, but how come
> the exact same kernel (the xenified one) could work fine without Xen
> loaded ? I am almost a noob in kernel/driver and stuff; but I thought
> the drivers were entirely in the kernel.
Yep, the driver is entirely in the kernel, but that's not the whole story.
Xen dom0 kernel does irq handling through Xen hypervisor,
so that might make some drivers behave in a different way baremetal vs. dom0.
Also remember dom0 is a *vm*, so some timing stuff might happen
differently on baremetal vs. dom0.
> I will try with the latest kernel in a few days.
> SLES11SP1 ships mptfusion 4.22
> I dont know for RHEL
What driver version does the squeeze kernel have?
> On Sat, Jan 29, 2011 at 6:02 PM, Pasi Kärkkäinen <pasik@xxxxxx> wrote:
> > On Sat, Jan 29, 2011 at 04:27:25PM +0100, Jordan Pittier wrote:
> >> Hi,
> >> I have been tracking a bug affecting all my servers running Debian
> >> Squeeze
> >> for more than a month now, and I*desperately*need your help :)*
> >> I have 10 Sun v20z servers (2*66GB SCSI disk in RAID 1 == mirror). 4 of
> >> them are running Debian Squeeze with the latest Xen Debian kernel
> >> (2.6.32-5-xen-amd64 ==*2.6.32-29). The rest are running Debian Lenny
> >> (2.6.26-2-xen-amd64 ==*2.6.26-26lenny1).
> >> On a Squeeze boxe, under very high IO (such as running a IO stress test,
> >> ie bonnie++), server starts behaving*weirdly and I see messages like
> >> these
> >> in kernel.log : [see attachement]. Then the server becomes totally
> >> unresponsive (but doesn't "freeze") and commands such as "ls" or
> >> "reboot"
> >> don't work anymore. I have to do an hard reboot. After the server has
> >> reboot, the RAID array seems degraded (I am using the mpt-status
> >> command)
> >> and starts rebuilding. After several hours, the raid array is "fine"
> >> ("clean"). The raid controler is "LSI53C1030" U320, with driver "Fusion
> >> MPT SPI Host driver 3.04.06". I have attached the result of "lsmod".
> >> None of my Lenny boxes are affected by this issue, all of my Squeeze
> >> boxes
> >> are.
> >> What does it have to do with Xen ? When I boot my Squeeze boxes without
> >> the Xen hypervisor but the same Xen kernel, bonnie++
> >> runs*absolutely*fine.
> >> The issue appears only with the Xen hypervisor loaded.*
> >> There is a debian bug report for this
> >> :*http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=603727
> >> Any suggestion ?*
> > Did you check if LSI has newer driver version available?
> > Also you might check which driver version for example RHEL6
> > or SLES11SP1 ships with.. both of those distros have 2.6.32 kernels too.
> > On one of my testboxes I need to upgrade the LSI driver
> > to a newer version to make it work. This is SAS based LSI though.
> > Can you try using another disk controller?
> > Also: Did you try using the latest kernel (-30) ?
> > -- Pasi
Xen-users mailing list