- 1 Virtual Machine Lifecycle
- 2 Concepts
- 3 Tasks
- 3.1 Creating a domain
- 3.2 Editing a domain
- 3.3 Starting a domain
- 3.4 Stopping or rebooting a domain
- 3.5 Pausing a guest domain
- 3.6 Unpausing a guest domain
- 3.7 Taking a Snapshot of a guest domain
- 3.8 Listing the snapshots of a guest domain
- 3.9 Restoring a guest domain from a snapshot
- 3.10 Removing a snapshot from a guest domain
- 3.11 Migration
- 3.12 Deletion of a domain
- 3.13 Wiping the storage used by a guest domain
- 4 Reference
Virtual Machine Lifecycle
This page describes the basics of the virtual machine lifecycle. Its aim is to provide fundamental information to create, run, stop, migrate and delete a virtual machines in one page.
It is always important to know what is meant by the terms used in documentation and meaning of commands and syntax. Please refer to this page if you are not familiar with basic libvirt concepts such as nodes and domains and to get an overview of the libvirt project goals and scope.
Guest domains are described using XML configuration
XML is used as the file format for storing the configuration of everything in libvirt including domain, network, storage and other elements. XML enables users to use the editors they are comfortable with and to easily integrate with other technologies and tools.
For example, devices in a domain are represented by XML elements by assigning attributes and/or child elements. A fragment is shown below:
<domain type='qemu'> <name>demo</name> ... <devices> ... <disk type='file' device='disk'> ... </disk> <disk type='file' device='cdrom'> ... </disk> <input type='mouse' bus='ps2'/> ... </devices> </domain>
Libvirt uses XPath technology to select nodes from XML document.
Transient guest domains vs Persistent guest domains
Libvirt distinguishes between two different types of domains: transient and persistent.
- Transient domains only exist until the domain is shutdown or when the host server is restarted.
- Persistent domains last indefinitely.
Once a domain is created (no matter what type) its state can be saved into a file and then restored indefinitely as long as the original file still exists. Thus even a transient domain can be restored over and over again.
Creation of transient domains differs slightly from the creation of persistent domains. Persistent domains need to be defined before being started up. Transient domains are created and started at once. The commands differ when dealing with the two different types.
While transient domains are created and destroyed on-the-fly, all of their components (e.g. storage, networks, devices, etc.) need to exist beforehand.
States that a guest domain can be in
Domain can be in several states:
- Undefined - This is a baseline state. Libvirt does not know anything about domains in this state because the domain hasn't been defined or created yet.
- Defined or Stopped - The domain has been defined, but it's not running. This state is also called stopped. Only persistent domains can be in this state. When a transient domain is stopped or shutdown, it ceases to exist.
- Running - The domain has been created and started either as transient or persistent domain. Either domain in this state is being actively executed on the node's hypervisor.
- Paused - The domain execution on hypervisor has been suspended. Its state has been temporarily stored until it is resumed. The domain does not have any knowledge whether it was paused or not. If you are familiar with processes in operating systems, this is the similar.
- Saved - Similar to the paused state, but the domain state is stored to persistent storage. Again, the domain in this state can be restored and it does not notice that any time has passed.
The diagram below shows how domain states flow into one another. The rectangles represent the different domain states and the arrows show the commands that move a domain from one state to another.
From the picture above, one can see that with the shutdown command one can move from a running state to a defined or undefined state. In this case, a transient domain would become undefined (cease to exist) and a persistent domain would be become defined upon a shutdown.
A snapshot is a view of a virtual machine's operating system and all its applications at a given point in time. Having a restorable snapshot of a virtual machine is a basic feature of the virtualization landscape. Snapshots allow users to save their virtual machine's state at a point in time and roll back to that state. Basic use cases include taking a snapshot, installing new applications, updates or upgrades (discovering they are terrible or broke things) and then rolling back to a prior time.
It should be obvious that any changes that occur after a snapshot is taken are not included in the snapshot. A snapshot does not continually update. It represents the virtual machine's state at a single point in time.
A running domain or virtual machine can be migrated to another host as long as the virtual machine's storage is shared between the hosts and the host's CPU is capable of supporting the guest's CPU model. Depending on the type and application, migration of virtual machines does not need to cause any service interruption.
Libvirt supports a number of different migration types:
- Standard- A domain is suspended while its resources are being transferred to the destination host. Once done, the VM resumes operation on the destination host. The time spent in suspended state is directly proportional to domain's memory size.
- Peer-to-peer - This type is used whenever source and destination hosts can communicate directly.
- Tunnelled - A tunnel is created between source and destination hosts, such as a SSH tunnel. All network communication between the source and destination nodes or physical hosts is sent through the tunnel.
- Live vs non-live - When migrating in live mode, the domain is not paused and all services on it continue to run. On the destination host the non-live domain or virtual machine has all processes stopped. The domain is effectively invisible during the time necessary to transfer its state through network. Live migration is therefore sensitive to application load. When live migrating a domain, it's allocated memory is sent to the destination host while being watched for changes on the source host. The domain on the source host remains active until all of the memory on both nodes are identical. At that point the domain on the destination node becomes active and the domain on the source node becomes passive or invisible to other machines on the network.
- Direct - libvirt initiates the migration using the hypervisor and then the process is entirely under control of the hypervisors. Often they have features to talk directly to each other (e.g. Xen on the source host communicates directly to Xen on the destination host without any libvirt intervention).
Requirements for migration:
- Shared storage accessible under same paths and locations,e.g. iSCSI, NFS, etc.
- Exactly the same versions of hypervisor software on both physical hosts
- Same network configuration.
- Same CPUs or better. The CPUs must be from the same vendor and the CPU flags on the destination host must be superset of CPU flags on the source host.
A checklist for a successful migration can be found here.
Guest data security when removing a domain
Some applications store sensitive information. As with any process that involves sensitive data, thought should be given to the safe and secure disposal of that information. Like any file on the filesystem, when a virtual machine is deleted from the system only the filesystem pointers are deleted. The occupied blocks on the storage media typically remain occupied, they are simply flagged as empty by the filesystem. Really, it depends on your filesystem.
Hopefully, if the application is dealing with such heavy data, then the machines themselves are physically secured and access to the network is similarly safeguarded.
Security is always an important factor to consider, even if only to keep out vandals and prevent innocent, but disastrous accidents.
Creating a domain
In order to run a domain it is first necessary to create one. This can be done in several ways. The following page describes the process using the Virtual Machine Manager GUI. The second way is by using the virt-install command line tool.
# virt-install \ --connect qemu:///system \ --virt-type kvm \ --name MyNewVM \ --ram 512 \ --disk path=/var/lib/libvirt/images/MyNewVM.img,size=8 \ --vnc \ --cdrom /var/lib/libvirt/images/Fedora-14-x86_64-Live-KDE.iso \ --network network=default,mac=52:54:00:9c:94:3b \ --os-variant fedora14
This command creates a new domain called 'MyNewVM', with 512 MB RAM and 8 GB disk space using KVM. Please read the manual page for any further information.
The last way is to create an XML definition of the domain and volume(s) and run virsh with the appropriate commands: vol-create and define.
Volumes are joined in a pool. By default, there exists one pool called "default". This is a directory-type pool, which means all volumes are stored as files in one directory. But please read this page if you are not completely aware of libvirt storage management. You may find more suitable storage solution there.
Example of volume XML definition (new_volume.xml):
<volume> <name>sparse.img</name> <capacity unit="G">10</capacity> </volume>
This defines a new volume with a capacity of 10 GB. To create volume in "default" pool:
# virsh vol-create default new_volume.xml
Example of domain XML definition (MyNewVM.xml):
<domain type='kvm'> <name>MyNewVM</name> <currentMemory>524288</currentMemory> <memory>524288</memory> <uuid>30d18a08-d6d8-d5d4-f675-8c42c11d6c62</uuid> <os> <type arch='x86_64'>hvm</type> <boot dev='hd'/> </os> <features> <acpi/><apic/><pae/> </features> <clock offset="utc"/> <on_poweroff>destroy</on_poweroff> <on_reboot>restart</on_reboot> <on_crash>restart</on_crash> <vcpu>1</vcpu> <devices> <emulator>/usr/bin/qemu-kvm</emulator> <disk type='file' device='disk'> <driver name='qemu' type='raw'/> <source file='/var/lib/libvirt/images/MyNewVM.img'/> <target dev='vda' bus='virtio'/> </disk> <disk type='block' device='cdrom'> <target dev='hdc' bus='ide'/> <readonly/> </disk> <interface type='network'> <source network='default'/> <mac address='52:54:00:9c:94:3b'/> <model type='virtio'/> </interface> <input type='tablet' bus='usb'/> <graphics type='vnc' port='-1'/> <console type='pty'/> <sound model='ac97'/> <video> <model type='cirrus'/> </video> </devices> </domain>
To define a new presistent domain:
# virsh define MyNewVM.xml
Domain XML format has many optional elements which you may find useful. Therefore read this page which is a complete domain XML format reference including examples and most common scenarios.
Editing a domain
Any domain can be edited in a user's favourite editor. What is needed is to set the $VISUAL or $EDITOR environment variable and run:
# virsh edit <domain>
If neither of these variables are not set, the vi editor is used by default. After closing the editor libvirt will automatically check for changes and apply them. However, it is also possible to edit domain in Virtual Machine Manager.
Starting a domain
Once a domain is created, one is able to run it. This is possible through Virtual Machine Manager or by running virsh start <domain> command. For example:
# virsh start MyNewVM
This command however performs either so called clean boot up or restores the domain from the previously saved state. See managedsave virsh command for details. It is important to notice, a domain can't be started if any of its components are not up, e.g. network.
As mentioned above, a transient domain can be run without previous definition:
# virsh create /path/to/MyNewVM.xml
Stopping or rebooting a domain
To stop running domain just run:
# virsh shutdown <domain>
To reboot a persistent domain:
# virsh reboot <domain>
Rebooting a transient domain is not possible, since right after shutdown are transient domains also undefined.
An inelegant shutdown, also known as hard-stop:
# virsh destroy <domain>
This is equivalent to unplugging the power cable.
Pausing a guest domain
Domain can be paused in virsh:
# virsh suspend <domain>
or in Virtual Machine Manager by clicking Pause button from main toolbar. When a guest is in a suspended state, it consumes system RAM but not processor resources. Disk and network I/O does not occur while the guest is suspended. This operation is immediate
Unpausing a guest domain
Any paused or suspended domain can be resumed by:
# virsh resume <domain>
or by unclicking the appropriate Pause button in Virtual Manager.
Taking a Snapshot of a guest domain
Creating a snapshot is done by executing:
# virsh snapshot-create <domain>
Listing the snapshots of a guest domain
All snapshosts of a guest domain can be viewed in virsh:
# virsh snapshot-list <domain>
For instance, the output might look like this:
Name Creation Time State --------------------------------------------------- 1295973577 2011-01-25 17:39:37 +0100 running 1295978837 2011-01-25 19:07:17 +0100 running
We can see one snapshot created at 17:39:17 local time, with the name 1295973577 which corresponds to Unix time. The other was created at 19:07:17 with the name 1295978837.
Restoring a guest domain from a snapshot
To restore a guest domain from a previous snapshot you can use:
# virsh snapshot-restore <domain> <snapshotname>
This restores a specified domain to a state represented by snapshotname. Please note that any changes made will be destroyed!
Removing a snapshot from a guest domain
Any snaphsot of a given domain can be removed via:
# virsh snapshot-delete <domain> <snapshotname>
Libvirt provides migration support. It means you can migrate a domain from one host to another over the network. Migration can operate in two main modes:
- Plain migration: The source host VM opens a direct unencrypted TCP connection to the destination host for sending the migration data. Unless a port is manually specified, libvirt will choose a migration port in the range 49152-49215, which will need to be open in the firewall on the remote host.
- Tunneled migration: The source host libvirtd opens a direct connection to the destination host libvirtd for sending migration data. This allows the option of encrypting the data stream. This mode doesn't require any extra firewall configuration, but is only supported with qemu 0.12.0 or later, and libvirt 0.7.2 or later.
For a successful migration there are couple of things needed to be done. For instance, storage settings have to match. All volumes that migrated domain use have to be stored under the same paths. For full checklist follow this page. We also recommend you use secure migration.
Once pre-migration checks are done, you can migrate machine using virsh:
# virsh migrate <domain> <remote host URI> --migrateuri tcp://<remote host>:<port>
Deletion of a domain
One may delete an inactive domain in virsh:
# virsh undefine <domain>
As usual, there is also the possibility of deleting it in Virtual Machine Manager, covered in this page .
Wiping the storage used by a guest domain
A volume used by a domain can contain confidential data, hence it is necessary to wipe it before removal. Libvirt offers a helping hand for such cases:
# virsh vol-wipe <volume>
which truncates and extends the volume to its original size. This in fact fills the file with zeroes. This ensures that data previously stored on volume is not accessible to reads anymore. After this, you can remove volume :
# virsh vol-delete <volume>
These pages may also provide useful further information: