Compute nodes start-up after ungraceful shutdown

From PRAGMA wiki
Jump to: navigation, search

When the VM hosting server went down ungracefully, the VM compute nodes (hosted-VM-x-x-x) will attempt to reinstall itself on start, eventhough the boot option for the node is set to "os".
To change this behavior, either turns off the install option using rocks-grub, or ssh to the corresponding vm-container and modify grub.conf on the compute node disk image.
For example, to fix the problem for a compute node hosted-vm-0-0-0 in vm-container-0-0 (in this case, the compute node /boot/grube/grub-orig.conf

# ssh vm-container-0-0
# cd /etc/xen/rocks
# ls
# cat *-0-0-0
# cd /state/partition1/xen/disks/
# ls -l
# df
# lomount -diskimage hosted-vm-0-0-0.hda -partition 1 /media/
# cd /media/boot/grube
# ls *conf
# more grub-orig.conf 
# cp grub-orig.conf grub.conf
# cd 
# umount /media

Then back to the VM hosting server and start the compute node (hosted-VM-x-x-x)