vm-control

README

vm-control

A utility to manage Virtual Machines (VMs).

The vm-control utility creates and manages VMs by communicating with a Hypervisor. It is typically run on a desktop, bastion or build machine.

Usage

vm-control supports several sub-commands. There are many command-line flags which provide parameters for these sub-commands. The most commonly used parameters are -fleetManagerHostname or -hypervisorHostname which specify either the Fleet Manager or a specific Hypervisor to communicate with. The basic usage pattern is:

vm-control [flags...] command [args...]

Built-in help is available with the command:

vm-control -h

Some of the sub-commands available are:

• add-vm-volumes: add volumes to a VM
• become-primary-vm-owner: become the primary owner of a VM
• change-vm-console-type: change the console type for a VM
• change-vm-cpus: change the number of CPUs for a VM
• change-vm-destroy-protection: enable/disable destroy protect for a VM
• change-vm-memory: change the memory for a VM
• change-vm-owner-users: change the extra owners for a VM
• change-vm-tags: change the tags for a VM
• change-vm-vcpus: change the number of vCPUs for a VM
• change-vm-volume-size: change the size of a specified volume on a VM
• connect-to-vm-console: connect to the Virtual Network Console for the specified VM
• connect-to-vm-serial-port: connect to the specified VM serial port
• copy-vm: make a copy of a VM. The new VM will have a different IP address
• create-vm: create a VM
• debug-vm-image: (re)start a VM with a temporary debug image. The old root volume will become the first secondary volume. Debugging ends when the VM is stopped or (re)started
• delete-vm-volume: delete a specified volume from a VM
• destroy-vm: destroy a VM (all ephemeral data and metadata are lost)
• discard-vm-old-image: discard the previous root image for a VM
• discard-vm-old-user-data: discard the previous user data for a VM
• discard-vm-snapshot: discard the previous snapshot for a VM
• export-local-vm: export a local VM to an importing tool. This is primarily for debugging
• export-virsh-vm: export VM to a local virsh VM. The specified FQDN will be used to specify the new virsh domain name. The VM must first be stopped. The exported virsh VM is started
• get-hypervisors: get details of healthy Hypervisors in the specified location
• get-vm-info: get and show the information for a VM
• get-vm-user-data: get (copy) the user data for a VM
• get-vm-volume: get (copy) a specified VM volume
• import-local-vm: import a local raw VM. This is primarily for debugging
• import-virsh-vm: import a local virsh VM. The specified domain name must be a FQDN, which is used to obtain the IP address of the imported VM. The virsh VM must first be shut down. The imported VM is started
• list-hypervisors: list healthy Hypervisors in the specified location
• list-locations: list locations within the specified top location
• list-vms: list the IP addresses for all VMs
• migrate-vm: migrate a VM to another Hypervisor
• patch-vm-image: patch the root image for a VM. Files listed in the image filter are not changed. The old root image is saved. The VM must not be running
• probe-vm-port: probe (from its Hypervisor) a TCP port for a VM
• reboot-vm: reboot a VM
• replace-vm-credentials: replace the credentials made available to a VM via the Metadata service
• replace-vm-image: replace the root image for a VM. The old root image is saved. The VM must not be running
• replace-vm-user-data: replace the user data for a VM. The old user data is saved
• restore-vm: restore all VM data (volumes) and metadata from a storage source. If the target Hypervisor has the original IP available it will be re-allocated for the new (restored) VM, otherwise a new IP address will be allocated
• restore-vm-from-snapshot: restore VM volumes from the previous snapshot, discarding current volumes
• restore-vm-image: restore the previously saved root image for a VM. The VM must not be running
• restore-vm-user-data: restore the previously saved user data for a VM
• reorder-vm-volumes: re-order the volumes for a VM
• set-vm-migrating: change the VM state to migrating. For debugging only
• snapshot-vm: create a snapshot of the VM volumes, discarding previous one
• save-vm: save (backup) all VM data (volumes) and metadata to a storage destination
• scan-vm-root: scan the root file-system of stopped VM and write to scanFilename
• start-vm: start a stopped VM
• stop-vm: stop a running VM. All data and metadata are preserved
• trace-vm-metadata: trace the requests a VM makes to the metadata service
• unset-vm-migrating: change the VM state to stopped. For debugging only

Security

The Hypervisor restricts RPC access using TLS client authentication. vm-control will load certificate and key files from the ~/.ssl directory. vm-control will present these certificates to the Hypervisor. If one of the certificates is signed by a certificate authority that the Hypervisor trusts, the Hypervisor will grant access. Most operations only require a certificate that proves identity. The Keymaster is a good choice for issuing these certificates.

Importing virsh (libvirt) VMs

A libvirt VM may be imported into the Hypervisor. Once the VM is committed it is removed from the libvirt database and is fully "owned" by the Hypervisor. Importing a VM requires root access on the Hypervisor (the vm-control tool will use the sudo command if needed).

There are a few simple steps that should be followed to import a VM. In the example below, the MAC address of the VM to be imported is 52:54:de:ad:be:ef and the hostname (DNS entry) is jump.prod.company.com. The IP address of the VM may also be used. In either case, the hostname or IP address provided must match the libvirt domain name. If the VM has multiple network interfaces, the MAC and IP address/FQDN for each interface must be provided in pairs.

• log into the VM and determine its MAC address
• run vm-control import-virsh-vm 52:54:de:ad:be:ef jump.prod.company.com
• enter shutdown at the prompt
• wait for the VM to start and log in and check that it is functioning properly
• respond to the commit/defer/abandon prompt:
  • commit: the VM is removed from the libvirt database
  • defer: the VM is left running on the Hypervisor. It may later be committed (and the virsh undefine command should be used to remove it from the libvirt database) or destroyed. This is not recommended
  • abandon: the VM is deleted from the Hypervisor and will need to be manually started with the virsh command

Exporting VMs to virsh (libvirt)

A local VM on the Hypervisor may be exported to a libvirt VM. Once the libvirt VM is committed the original VM is removed from the database and is fully "owned" by libvirt. Exporting a VM requires root access on the Hypervisor (the vm-control tool will use the sudo command if needed).

There are a few simple steps that should be followed to export a VM. In the example below, the hostname (DNS entry) of the VM to be exported is jump.prod.company.com. The IP address of the VM may also be used. In either case, the hostname or IP address provided will become the new libvirt domain name.

• set up a DHCP server
• add the IP and MAC addresses of the VM being exported to the DHCP server configuration. Alternatively, you can log into the VM prior to exporting it and reconfigure it to use a static network configuration
• if a Fleet Manager is being used, the IP address must be added to the ReservedIPs list in the topology, so as to prevent re-use of the IP address when creating new VMs
• run vm-control export-virsh-vm jump.prod.company.com
• you will be prompted if a DHCP server entry has been configured
• wait for the VM to start and log in and check that it is functioning properly
• respond to the commit/defer/abandon prompt:
  • commit: the VM is removed from the Hypervisor database
  • defer: the new libvirt VM is left running. It may later be committed (and the vm-control destroy-vm command should be used to remove it from the Hypervisor database). Deferring is not recommended
  • abandon: the new libvirt VM is deleted from the libvirt database and the original VM will be started

VM Placement Command

An optional local command to be used when making VM placement decisions (when creating, copying, migrating or restoring VMs) may be specified using the -placementCommand option when -placement=command. The placement command may read from the standard input a JSON payload describing the Hypervisors available (based on other location and tag matching options) and the VmInfo describing the VM that will be placed. The name of the selected Hypervisor must be written to the standard output.

The schema of the data presented to the standard input is:

struct {
    Hypervisors []fm_proto.Hypervisor `json:",omitempty"`
    VmInfo      hyper_proto.VmInfo
}

See fm_proto and hyper_proto for sub-schema definitions.