I’ve been using virt-manager to manage my KVM hosts and I’m not keen on having to login to the remote hosts as root, plus I would get the password prompt every time I connect to the server (sure I could setup my pulic SSH key on the root account, but not a good idea to use RSA auth to the root account on a remote server).  With Debian (Wheezy) it was fairly simple in that all that I had to do was add my regular username to the group “libvirt”.  Then I could use the URI: qemu+ssh://virtadmin@my.kvmhost.com/system to connect to the remote KVM host using virt-manager.

I still had trouble doing things like migration and I couldn’t use virsh from the CLI.  Not the the end of the world, just figured it was a bug or oddity because I am using Fedora as my laptop distro and Debian on the KVM host.  But I could use virt-manager to create and manage my guests remotely.

As usual, I get itchy and want to try new distros and such so I wanted to see what KVM was like under Centos 6 since that’s what I use at work.  So I migrated all of my guests over to one KVM host, then installed CentOS 6.5 minimal on the other server and installed all of the necessary KVM and libvirt RPMs.

$ yum install libvirt libvirt-client libvirt-python qemu-kvm libguestfs-tools

But I soon found out that no matter which group I added my username to under CentOS, I couldn’t do anything as a normal user, either from virt-manager or the CLI on the server.  So after some Googlin’ I came across the libvirt wiki page outlining how to use PolicyKit to allow normal users to access and use libvirt.  Cool!  And thankfully since CentOS 6.5 still uses PolKit 0.95 I don’t have to create the file in JavaScript!  So I followed the instructions, added my username to the libvirt group and viola! I could now access the remote server from virt-manager.  However, I still couldn’t do anything using virsh from the CLI on the host server.  Turns out you need to specify the URI when using virsh even though you are part of a group that has full access to libvirt.  Using this post, I added the necessary stanza to .bash_profile and was off and running:

$ vi ~/.bash_profile

# the following stanza will allow the use of virsh without having
# having to specify the qemu+ssh URI every time
if test -x `which virsh`; then
  export LIBVIRT_DEFAULT_URI=qemu:///system
fi

$ source ~/.bash_profile

SELinux Note: If you keep SELinux on (SELINUX=enforcing in /etc/selinux/config), then you will need to follow the instructions on this post to make sure that libvirt can properly access resources on folders, NFS mounts, etc. that aren’t part of the default locations setup when you install KVM/qemu/libvirt.  Me, I’ve set it to SELINUX=permissive so I don’t have to constantly guess why I get errors when trying to do something. Gotta love SELinux.  Maybe someone can explain why I should keep it on…

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s