Connection URIs

Since libvirt supports many different kinds of virtualization (often referred to as "drivers" or "hypervisors"), we need a way to be able to specify which driver a connection refers to. Additionally we may want to refer to a driver on a remote machine over the network.

To this end, libvirt uses URIs as used on the Web and as defined in RFC 2396. This page documents libvirt URIs.

Specifying URIs to libvirt

The URI is passed as the name parameter to virConnectOpen or virConnectOpenReadOnly. For example: 

virConnectPtr conn = virConnectOpenReadOnly ("test:///default");

Specifying URIs to virsh, virt-manager and virt-install

In virsh use the -c or --connect option: 

virsh -c test:///default list

If virsh finds the environment variable VIRSH_DEFAULT_CONNECT_URI set, it will try this URI by default.

When using the interactive virsh shell, you can also use the connect URI command to reconnect to another hypervisor.

In virt-manager use the -c or --connect=URI option: 

virt-manager -c test:///default

In virt-install use the --connect=URI option: 

virt-install --connect=test:///default [other options]

xen:/// URI

This section describes a feature which is new in libvirt > 0.2.3. For libvirt ≤ 0.2.3 use "xen".

To access a Xen hypervisor running on the local machine use the URI xen:///.

qemu:///... QEMU and KVM URIs

To use QEMU support in libvirt you must be running the libvirtd daemon (named libvirt_qemud in releases prior to 0.3.0). The purpose of this daemon is to manage qemu instances.

The libvirtd daemon should be started by the init scripts when the machine boots. It should appear as a process libvirtd --daemon running as root in the background and will handle qemu instances on behalf of all users of the machine (among other things).

So to connect to the daemon, one of two different URIs is used:

(If you do libvirtd --help, the daemon will print out the paths of the Unix domain socket(s) that it listens on in the various different modes).

KVM URIs are identical. You select between qemu, qemu accelerated and KVM guests in the guest XML as described here.

Remote URIs

Remote URIs are formed by taking ordinary local URIs and adding a hostname and/or transport name. As a special case, using a URI scheme of 'remote', will tell the remote libvirtd server to probe for the optimal hypervisor driver. This is equivalent to passing a NULL URI for a local connection. For example:

Local URI Remote URI Meaning
xen:/// xen://oirase/ Connect to the Xen hypervisor running on host oirase using TLS.
NULL remote://oirase/ Connect to the "default" hypervisor running on host oirase using TLS.
xen:/// xen+ssh://oirase/ Connect to the Xen hypervisor running on host oirase by going over an ssh connection.
test:///default test+tcp://oirase/default Connect to the test driver on host oirase using an unsecured TCP connection.

Remote URIs in libvirt offer a rich syntax and many features. We refer you to the libvirt remote URI reference and full documentation for libvirt remote support.

test:///... Test URIs

The test driver is a dummy hypervisor for test purposes. The URIs supported are:

Other & legacy URI formats

NULL and empty string URIs

Libvirt allows you to pass a NULL pointer to virConnectOpen*. Empty string ("") acts in the same way. Traditionally this has meant connect to the local Xen hypervisor. However in future this may change to mean connect to the best available hypervisor.

The theory is that if, for example, Xen is unavailable but the machine is running an OpenVZ kernel, then we should not try to connect to the Xen hypervisor since that is obviously the wrong thing to do.

In any case applications linked to libvirt can continue to pass NULL as a default choice, but should always allow the user to override the URI, either by constructing one or by allowing the user to type a URI in directly (if that is appropriate). If your application wishes to connect specifically to a Xen hypervisor, then for future proofing it should choose a full xen:/// URI.

File paths (xend-unix-server)

If XenD is running and configured in /etc/xen/xend-config.sxp: 

(xend-unix-server yes)

then it listens on a Unix domain socket, usually at /var/lib/xend/xend-socket. You may pass a different path using a file URI such as: 

virsh -c ///var/run/xend/xend-socket

Legacy: http://... (xend-http-server)

If XenD is running and configured in /etc/xen/xend-config.sxp: 

(xend-http-server yes)

then it listens on TCP port 8000. libvirt allows you to try to connect to xend running on remote machines by passing http://hostname[:port]/, for example: 

virsh -c http://oirase/ list

This method is unencrypted and insecure and is definitely not recommended for production use. Instead use libvirt's remote support.

Notes:

  1. The HTTP client does not fully support IPv6.
  2. Many features do not work as expected across HTTP connections, in particular, virConnectGetCapabilities. The remote support however does work correctly.
  3. XenD's new-style XMLRPC interface is not supported by libvirt, only the old-style sexpr interface known in the Xen documentation as "unix server" or "http server".

Legacy: "xen"

Another legacy URI is to specify name as the string "xen". This will continue to refer to the Xen hypervisor. However you should prefer a full xen:/// URI in all future code.

Legacy: Xen proxy

Libvirt continues to support connections to a separately running Xen proxy daemon. This provides a way to allow non-root users to make a safe (read-only) subset of queries to the hypervisor.

There is no specific "Xen proxy" URI. However if a Xen URI of any of the ordinary or legacy forms is used (eg. NULL, "", "xen", ...) which fails, and the user is not root, and the Xen proxy socket can be connected to (/tmp/libvirt_proxy_conn), then libvirt will use a proxy connection.

You should consider using libvirt remote support in future.