With very many thanks to Jean-Dominique Leca how mailed me with the following that worked
============================ Hello, This is "fixed". If you're using an i686 kernel, you have to rebuild the Hyper-V components against the 2.6.32-279.11.1.el6 kernel. Don't know why it wasn't necessary with the x86_64 kernel. prerequisites : - remove the hyper-v components - boot using the latest kernel (2.6.32-279.11.1.el6) - if necessary, add a legacy network card for remote access (facilitates operations). Steps to rebuild : - setup a working rpmbuild infrastructure (http://wiki.centos.org/HowTos/SetupRpmBuildEnvironment, http://wiki.centos.org/HowTos/RebuildSRPM) - install the kernel-devel package - from the LinuxICv34 iso and as a normal user, install the src.rpm package (rpm -ivh microsoft-hyper-v-rhel63.3.4-1.20120727.src.rpm) - this will install an hyper-v.spec file into your ~/rpmbuild/SPECS directory - as a normal user, cd to ~/rpmbuild/SPECS and run 'rpmbuild -bb microsoft-hyper-v.spec' - this will compile the hyper-v components agains your current kernel and build the related rpm packages. - install the new RPMs, first kmod-microsoft-hyper-v and after microsoft-hyper-v (rpm -ivh .). The issue is related to the ata_piix driver, when the hyper-v components are working, this driver is removed and replaced with the hv_storvsc driver (/etc/modprobe.d/hyperv_pvdrivers.conf : install ata_piix { /sbin/modprobe hv_storvsc 2>&1 || /sbin/modprobe --ignore-install ata_piix; } I wasted some time because my first approach was to patch the ata_piix driver, but it turns out to be quite unsuccessful (i even try to backport some code from the 3.6 branch, guess what happened.). So the obvious fix is to rebuild the hyper-v components. ============================ -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Jason Ozin Sent: 28 October 2012 12:46 To: 'BlueOnyx General Mailing List' Subject: [BlueOnyx:11641] Re: Serious issues with Kernel 2.6.32-279.11.1.el6 Following my request below. Can anyone point me in the right direction to help me debug this? I'm concerned that once the kernel upgrade bug is fixed I will be forced onto the new kernel and have two broken servers. Thanks Jason -----Original Message----- Sent: 21 October 2012 15:58 To: [email protected] Subject: [BlueOnyx:11567] Serious issues with Kernel 2.6.32-279.11.1.el6 Serious issues with Kernel 2.6.32-279.11.1.el6 on two different servers If anyone can help I would be most appreciative. Currently running: 5107R Kernel Version: 2.6.32-279.9.1.el6.i686 (SMP) i686 As a Hyper-V virtual guest with MS Hyper-V components installed (Linux Integration Services Version 3.4 for Hyper-V) At the moment 5107R still has the kernel upgrade bug where a kernel upgrade does not take effect unless you manually change /boot/grub/grub.conf So I noticed that Kernel 2.6.32-279.11.1.el6 had been installed and thus I edited /boot/grub/grub.conf in 3 places to use this new Kernel. Booting into the Kernel works and the server becomes responsive on the network for a few seconds then something major crashes and takes the server down. The console shows all sorts of strange output but the last two lines are: [<c052e0a1>] ? sys_read+0x41/0x70 [<c0409a9f>] ? sysenter_do_call+0x12/0x28 A screenshot of the console output is here: http://www2.ozin.com/boot.png I can't see anything particular in dmesg but if you want to take a look it is here: http://www2.ozin.com/dmesg.txt Editing /boot/grub/grub.conf to use 2.6.32-279.9.1.el6.i686 (SMP) i686 fixes the issue so it seems to be something about this new Kernel. Any ideas where I should start looking? Thanks in advance Jason _______________________________________________ Blueonyx mailing list [email protected] http://mail.blueonyx.it/mailman/listinfo/blueonyx _______________________________________________ Blueonyx mailing list [email protected] http://mail.blueonyx.it/mailman/listinfo/blueonyx
