Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

I setup my kvm virtualization environment with GlusterFS, the topology as follows:

KVM1 ---------- KVM2 --------------KVM3

|----------------------|-------------------------|

-----------------------|-------------------------

--------------|------------------|----------------

-------GlusterFS1---GlusterFS2----------

the description of setup as follows:

Mount GlusterFS

three kvm machines mount glusterfs with command:

/sbin/mount -t glusterfs -o sync,direct-io-mode=enable gfs1.dalegames.com:/gfs /gfs

Creating file type disk for virtual machine

qemu-img create -f qcow2 /gfs/vs_data/linux 100G #For Linux virtual machine

qemu-img create -f qcow2 /gfs/vs_data/windows 100G #For Windows virtual machine

Installing Virtual Machine

Installing Linux Virtual Machine

virt-install --ram=4096 --vcpus=4 --os-type=linux --os-variant=rhel6 \ --cdrom /gfs/ISOs/RHEL_Server_6.5_x86_64_DaleVS.iso --network bridge=br0 \ --disk path=/gfs/vs_data/linux,format=qcow2 --accelerate --vnc \ --name=linux --noautoconsole

Installing Windows Virtual Machine

virt-install --ram=4096 --vcpus=4 --os-type windows --os-variant=win2k8 \ --cdrom /gfs/ISOs/Win2k8.iso --network bridge=br0 \ --disk path=/gfs/vs_data/windows,format=qcow2,bus=virtio --disk /gfs/ISOs/virtio-win-disk.vfd,device=floppy\ --accelerate --vnc \ --name=windows --noautoconsole --hvm

Live Migrating Virtual Machine and the Question

[root@kvm1 ~]# virsh migrate --live --verbose linux qemu+ssh://kvm2/system

root@kvm2's password:

Migration: [100 %]

At this time, the file-system of linux virtual machine went to read-olny mode.

But when I live migrate a windows virtual machine, all works well.

Anyone can help me to fix it. thanks a lot!

share|improve this question
    
the error messages: Message from syslogd@newvs at Jan 8 22:33:01 ... kernel:journal commit I/O error –  user204702 Jan 8 at 6:34
    
Looks like a storage level issue, if the VM didn't get paused, that means qemu io calls went through, but the guest OS still felt a glitch, hence the switch to r/o. check the storage level logs for this time on both hosts –  dyasny Jan 9 at 4:10
    
What is the ownership on your images after migration? Apparently, Libvirt doesn't detect glusterfs as a shared filesystem like it does NFS or Samba, so when a VM migrates away, the original host will chown root:root the disk image. Since the receiving host is running the VM as qemu:qemu (on RHEL6/CentOS6, at least), the image on glusterfs is now read-only to the VM. See: bugzilla.redhat.com/show_bug.cgi?id=1057645#c7 –  Mark R Feb 26 at 15:10

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.