This page describes the main principles and architecture choices behind the definition of the libvirt API:
As defined in the goals section, the libvirt
API is designed to expose all the resources needed to manage the
virtualization support of recent operating systems. The first object
manipulated through the API is the virConnectPtr, which
represents the connection to a hypervisor. Any application using libvirt
is likely to start using the
API by calling one of the virConnectOpen functions. You will note that those functions take
a name argument which is actually a connection URI
to select the right hypervisor to open.
A URI is needed to allow remote connections and also select between
different possible hypervisors. For example, on a Linux system it may be
possible to use both KVM and LinuxContainers on the same node. A NULL
name will default to a preselected hypervisor, but it's probably not a
wise thing to do in most cases. See the connection
URI page for a full descriptions of the values allowed.
OnDevice the application obtains a
virConnectPtr
connection to the hypervisor it can then use it to manage the hypervisor's
available domains and related virtualization
resources, such as storage and networking. All those are
exposed as first class objects and connected to the hypervisor connection
(and the node or cluster where it is available).
The figure above shows the five main objects exported by the API:
virConnectPtr
Represents the connection to a hypervisor. Use one of the virConnectOpen functions to obtain connection to the hypervisor which is then used as a parameter to other connection API's.
virDomainPtr
Represents one domain either active or defined (i.e. existing as
permanent config file and storage but not currently running on that
node). The function
virConnectListAllDomains
lists all the domains for the hypervisor.
virNetworkPtr
Represents one network either active or defined (i.e. existing
as permanent config file and storage but not currently activated).
The function
virConnectListAllNetworks
lists all the virtualization networks for the hypervisor.
virStorageVolPtr
Represents one storage volume generally used
as a block device available to one of the domains. The function
virStorageVolLookupByPath
finds the storage volume object based on its path on the node.
virStoragePoolPtr
Represents a storage pool, which is a logical area
used to allocate and store storage volumes. The function
virConnectListAllStoragePools
lists all of the virtualization storage pools on the hypervisor.
The function
virStoragePoolLookupByVolume
finds the storage pool containing a given storage volume.
Most objects manipulated by the library can also be represented using XML descriptions. This is used primarily to create those object, but is also helpful to modify or save their description back.
Domains, networks, and storage pools can be either active
i.e. either running or available for immediate use, or
defined in which case they are inactive but there is
a permanent definition available in the system for them. Based on this
they can be activated dynamically in order to be used.
Most objects can also be named in various ways:
name
A user friendly identifier but whose uniqueness cannot be guaranteed between two nodes.
ID
A runtime unique identifier provided by the hypervisor for one given activation of the object; however, it becomes invalid once the resource is deactivated.
UUID
A 16 byte unique identifier as defined in RFC 4122, which is guaranteed to be unique for long term usage and across a set of nodes.
The naming of the functions present in the library is usually composed by a prefix describing the object associated to the function and a verb describing the action on that object.
For each first class object you will find APIs for the following actions:
Used to perform lookups on objects by some type of identifier, such as:
Used to enumerate a set of object available to an given hypervisor connection such as:
Generic accessor providing a set of generic information about an object, such as:
Specific accessors used to query or modify data for the given object, such as:
Used to create and start objects. The ...CreateXML APIs will create the object based on an XML description, while the ...Create APIs will create the object based on existing object pointer, such as:
Used to shutdown or deactivate and destroy objects, such as:
Note: functions returning vir*Ptr (like the virDomainLookup functions) allocate memory which needs to be freed by the caller by the corresponding vir*Free function (e.g. virDomainFree for a virDomainPtr object).
For more in-depth details of the storage related APIs see the storage management page.
Drivers are the basic building block for libvirt functionality
to support the capability to handle specific hypervisor driver calls.
Drivers are discovered and registered during connection processing as
part of the
virInitialize
API. Each driver
has a registration API which loads up the driver specific function
references for the libvirt APIs to call. The following is a simplistic
view of the hypervisor driver mechanism. Consider the stacked list of
drivers as a series of modules that can be plugged into the architecture
depending on how libvirt is configured to be built.
The driver architecture is also used to support other virtualization components such as storage, storage pools, host device, networking, network interfaces, and network filters.
See the libvirt drivers page for more information on hypervisor and storage specific drivers.
Not all drivers support every virtualization function possible. The libvirt API support matrix lists the various functions and support found in each driver by the version support was added into libvirt.
Access to libvirt drivers is primarily handled by the libvirtd
daemon through the remote driver via an
RPC. Some hypervisors do support
client-side connections and responses, such as Test, OpenVZ, VMware,
Power VM (phyp), VirtualBox (vbox), ESX, Hyper-V, Xen, and Parallels.
The libvirtd daemon service is started on the host at system boot
time and can also be restarted at any time by a properly privileged
user, such as root. The libvirtd daemon uses the same libvirt API
virInitialize
sequence as applications
for client-side driver registrations, but then extends the registered
driver list to encompass all known drivers supported for all driver
types supported on the host.
The libvirt client applications use a
URI to obtain the virConnectPtr.
The virConnectPtr keeps track of the driver connection
plus a variety of other connections (network, interface, storage, etc.).
The virConnectPtr is then used as a parameter to other
virtualization functions. Depending upon the
driver being used, calls will be routed through the remote driver to
the libvirtd daemon. The daemon will reference the connection specific
driver in order to retrieve the requested information and then pass
back status and/or data through the connection back to the application.
The application can then decide what to do with that data, such as
display, write log data, etc. Migration
is an example of many facets of the architecture in use.
The key takeaway from the above diagram is that there is a remote driver
which handles transactions for a majority of the drivers. The libvirtd
daemon running on the host will receive transaction requests from the
remote driver and will then query the hypervisor driver as specified in
the virConnectPtr in order to fetch the data. The data will
then be returned through the remote driver to the client application
for processing.
If you are interested in contributing to libvirt, read the FAQ and hacking guidelines to gain an understanding of basic rules and guidelines. In order to add new API functionality follow the instructions regarding implementing a new API in libvirt.