Latest Centos 6 kernel not booting.

KFC-Netearth

Hi,

I have just yum updated my Centos6 VMs and the new kernel:
2.6.32-754.24.2.el6.centos.plus.x86_64
does not boot.

On XCP-ng Center the VM goes green moves to a server and then shutdowns back to red.
The is virtualization state is: Paravirtualization (PV)

If I edit grub.conf and change back to the previous kernel it boots OK.
I am running XCP 8.0 with the latest updates.
Please let me know if you need any further info.

Regards
Kevin
Please let me know if you need any

maxcuttins

This is WEEEEEEEEEEEL know issue.
CentOS marked this bug as WON'T FIX!
So keep the previous kernel and that's it.

https://bugs.centos.org/view.php?id=15074

r1

You can change the VM type to HVM and it will boot fine. Do you have any particular reason to run it in PV?

KFC-Netearth

Thanks for the info. but I have booted this VM on kernel 6.32-754.2.1 and all since then.
I'll try setting eagerfpu=off

KFC-Netearth

@r1 Thanks - I'll try that, its a PV because that's how it was created by default back in 2012.

KFC-Netearth

Well eagerfpu=off makes no difference and converting to a HVM brings up the grub boot menu but the new kernel shuts down like before and the other kernels just hang, so the VM is now completely unbootable.
I'll try to convert back to a PV.
If that doesn't work, I have a pre-convert copy to boot from.

r1

Remove quite rhgb spalsh from kernel command line and add noreboot to see why its faulting.

olivierlambert

In general, I would suggest to go for HVM mode instead of PV.

KFC-Netearth

Converting back ti PV allow be to boot the old kernel again:
2.6.32-754.23.1.el6.centos.plus.x86_64

I'll change the new kernel command from:

kernel /vmlinuz-2.6.32-754.24.2.el6.centos.plus.x86_64 ro root=/dev/mapper/VolGroup-lv_root rd_NO_LUKS LANG=en_US.UTF-8 rd_NO_MD console=hvc0  KEYTABLE=us rd_LVM_LV=VolGroup/lv_swap SYSFONT=latarcyrheb-sun16 rhgb crashkernel=auto quiet  rd_LVM_LV=VolGroup/lv_root rd_NO_DM

to

kernel /vmlinuz-2.6.32-754.24.2.el6.centos.plus.x86_64 ro root=/dev/mapper/VolGroup-lv_root rd_NO_LUKS LANG=en_US.UTF-8 rd_NO_MD console=hvc0  KEYTABLE=us rd_LVM_LV=VolGroup/lv_swap SYSFONT=latarcyrheb-sun16 crashkernel=auto noreboot  rd_LVM_LV=VolGroup/lv_root rd_NO_DM

In PV mode all kernels shut down before the console appears.
IN HNV mode the new kernel displays:
Probing edd=off to disable
then shuts down after a few seconds and the older kernels shut down with just a blank screen.
Adding eff=off to the kernel line stops "Probing edd=off to disable"
appearing but the boot stops as before.

Maybe I'll try another Centos 6 VM to see if it does the same or if its just a problem with this VM.

r1

Your new kernel's initrd is missing Xen drivers. You will need to boot on a working kernel and then you can add those using # dracut --add-drivers "xen-blkfront xen-netfront" --kver 2.6.32-754.24.2.el6.centos.plus.x86_64 I'm not sure on 2.6.32-754.24.2.el6.centos.plus.x86_64 string but I guess it should work, else share # ll -ht /boot/.

HVM is always recommended than PV.. it will give you noticeable performance improvement.

KFC-Netearth

Yes that's it - thank you

I should have noticed the initramfs line was missing in grub.conf.
Strange, for some reason the new kernel install did not create a new initramfs file in /boot.
I've just updated an old Centos 6 PV VM and the kernel update creates the new initramfs file and boots the new kernel OK.
Maybe I'll remove the new kernel on the bad VM and install it again as there may be other problems with it.
The other 4 VMs I updated this morning have a initramfs file for the new kernel.

And I'll try HVM mode

r1

@KFC-Netearth Welcome