Issue
After upgrading compute resources to Xen >= 4.8, some virtual servers keep crashing or a kernel panic occurs in a few minutes after they are started.
Environment
OnApp 6.x with computer resources based on Xen >=4.8 compute resources
Some of the affected distributions and their kernel versions:
- Debian 8.x kernel version 3.16.*
- CentOS 7.x kernel kernel version 3.10.*
- Ubuntu 14.x kernel version 3.13.*
- Ubuntu 13.x kernel version 3.8.*
Resolution
1. Edit a guest's GRUB configuration file adding the mce=off boot option to the kernel parameters.
GRUB Legacy:
Edit the bootloader config file and append this option to the kernel line:
• /boot/grub/menu.lst on Debian
• /boot/grub/grub.conf on CentOS
GRUB2:
• Edit /etc/default/grub and append this option between the quotes in
the GRUB_CMDLINE_LINUX_DEFAULT line.
• Regenerate the grub.cfg file with grub2-mkconfig -o /boot/grub/grub.cfg
OR
• Edit /boot/grub/grub.cfg and append this option to the Linux line.
When the above is not possible to apply while the virtual server is running, boot it in recovery mode and perform these changes from there. Refer to the Recovery Console page.
If you want to have a VS's disk be mounted on computer resources, refer to the Mount Virtual Server's Disk to Compute Resource or Backup Server page.
If you want to run grub2-mkconfig -o /boot/grub/grub.cfg when the virtual server is in recovery mode, prepare the environment and chroot there. Refer to the Manually Reset Linux Root Password via Recovery page.
2. Make the changes at the compute resource level. These changes will be applied to all the Linux-based virtual servers that start on this compute resource.
Edit the pygrub bootloader script /usr/lib64/xen/bin/pygrub and add mce=off in the following line:
grubcfg = { "kernel": None, "ramdisk": None, "args": "mce=off " }
(! spare space "mce=off<SPACE>" does matter)
CODE
For more information about the issue, refer to: