hyper-control

README

hyper-control

A utility to manage SmallStack Hypervisors.

The hyper-control utility manages Hypervisors. It is typically run on a desktop or bastion machine. Please read the SmallStack design document to understand the architecture.

Usage

hyper-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. At startup, hyper-control will read parameters from the ~/.config/hyper-control/flags.default and ~/.config/hyper-control/flags.extra files. These are simple name=value pairs. The basic usage pattern is:

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

Built-in help is available with the command:

hyper-control -h

Some of the sub-commands available are:

• add-address: manually add a MAC address and IP address pair to a specific Hypervisor. If the IP address is not specified an external DHCP server is required to provides leases to VMs. This is only required if a Fleet Manager is not available
• add-subnet: manually add a subnet to a specific Hypervisor. This is only required if a Fleet Manager is not available
• change-tags: change the tags for a specific Hypervisor
• get-machine-info: get information for a specific Hypervisor
• get-updates: get and show a continuous stream of updates from a Hypervisor or Fleet Manager. This is primarily for debugging
• installer-shell: start a remote shell (via SRPC) to the installer running on a machine
• make-installer-iso: make a bootable installation ISO (CD-ROM) image for a machine
• move-ip-address: move a (free) IP address to a specific Hypervisor
• netboot-host: temporarily enable PXE-based network booting and installing for a machine
• netboot-machine: temporarily enable PXE-based network booting for a machine
• netboot-vm: create a temporary VM and install with PXE booting. This is for debugging physical machine installation
• power-off: shut down and power off the specified Hypervisor. All VMs must be stopped beforehand
• power-on: power on the specified Hypervisor. This uses remote IPMI or Wake On LAN, where available.
• register-external-leases: register external DHCP leases with a specific Hyervisor. These are lost after a Hypervisor restart
• reinstall: reinstall the local machine. This erases all data
• remove-excess-addresses: remove free addresses for a specific Hypervisor above the specified limit
• remove-ip-address: remove a (free) IP address from a specific Hypervisor
• remove-mac-address: remove a (free) MAC address from a specific Hypervisor
• rollout-image: safely roll out specified image to all Hypervisors in a location
• write-netboot-files: write the configuration files for installing a machine. This is primarily for debugging

Security

The Hypervisor restricts RPC access using TLS client authentication. hyper-control will load certificate and key files from the ~/.ssl directory. hyper-control will present these certificates to the Hypervisor (or Fleet Manager). If one of the certificates is signed by a certificate authority that the Hypervisor trusts, the Hypervisor will grant access.

Installing Hypervisors

The hyper-control tool may be used to install Hypervisors (OS+Hypervisor) on physical machines. This requires that information about machines and subnets is recorded in the topology (usually in Git), which is obtained from the Fleet Manager. Hypervisors may be installed by PXE booting an installer, booting from a custom ISO image or installing over a running system. Please read the Machine Birthing design document which describes the principles of installing physical machines.

Network (PXE) Installation

This is the most common method of installing. If there is at least one working Hypervisor on the same subnet, you can PXE (network) boot. The hyper-control tool is used to automatically select a Hypervisor to configure as a PXE server (it creates a temporary DHCP lease and will serve configuration files via TFTP).

The following options must be provided:

• fleetManagerHostname
• imageServerHostname
• installerImageStream

You may want to increase the -netbootTimeout option if the machine takes a long time to boot. Run the following command:

hyper-control netboot-host $target_host

Then, initiate a PXE boot for the target machine. It should boot the installer image, configure and install the machine and then reboot into the new OS. In principle, multiple machines can be installed in parallel, one machine for each time the above command is run.

Installing from an ISO (CD-ROM) image

If there is no working Hypervisor on the subnet and if there is no DHCP relay configured to forward DHCP requests to a Hypervisor on another subnet, then another option is to install the machine using a custom ISO (CD-ROM) image. The ISO image can be written to a CD-ROM, a USB memory drive or served by a NFS/SMB server (this requires that the machine BIOS supports booting from remote media).

As above, the same required options must be provided. The following command will generate a custom ISO image for the target machine. Note that the networking configuration for the machine is baked into the ISO, so the ISO is only good for one machine.

hyper-control make-installer-iso $target_host $destdir

This will create a custom ISO for this machine in the specified directory, with both hostname and IP address file entries.

Self installing a Hypervisor

If there is a Linux OS already installed and running on a machine, you can use the hyper-control utility to install a new OS+Hypervisor. This uses the kexec utility to boot directly into the new kernel, skipping the BIOS. This is a good way of wiping and getting a fresh install.

As above, the same required options must be provided. Run this command:

hyper-control reinstall

This method is much faster, since it skips rebooting via the BIOS the first. Remember: it wipes all data on the machine!

Upgrading Hypervisors

Upgrading Hypervisors is done using the hyper-control tool. This sets the RequiredImage tag on Hypervisors in the specified location. The rollout is controlled, starting slow and gaining speed as Hypervisors complete upgrades and remain healthy. The upgrade may include a new kernel, in which case the machine will be rebooted as part of the upgrade, thus taking several minutes for the per-machine upgrade+health check cycle to complete (compared to less than one minute for most upgrades). The actual upgrades are performed by the dominator which reads the tags to determine the image to push to the machine. Once the rollout is complete, the new tags are saved by committing to the Git repository containing the topology.

The following options must be provided:

• fleetManagerHostname
• imageServerHostname
• location
• topologyDir

To rollout a new Hypervisor image to all the Hypervisors in the specified location, run a command like this:

hyper-control rollout-image $image_name