Node OS Provisioning

ClusterVisor has a full featured set of operating system provisioning tools.

ClusterVisor has a full featured set of operating system provisioning tools. You can boot nodes up either stateful where a full Linux distros installed on the nodes, or stateless where disks are not used on the nodes, and the operating system in RAM only. ClusterVisor has tools to help you create these images, either from a Linux distribution installation from an ISO file, or by taking an image from an existing installed node.

ClusterVisor Supported Linux distros:

  • RedHat Enterprise Linux 8 & 9
  • Rocky Linux 8 & 9

Value of a ClusterVisor appliance

ClusterVisor can not only image compute nodes, but head nodes, login nodes, storage, etc. When you have a ClusterVisor appliance all the images are stored there. This makes it a great disaster recovery solution, as if there is ever a problem with a login or storage node, it can easily be re-imaged from the appliance, allowing you to get a critical down system up and running in no time.

Images from a Linux distribution

ClusterVisor allows you to upload a distribution ISO to the ClusterVisor appliance or server. Once uploaded, you can generate new operating system images from the uploaded distro. Roles of pre-defined software packages and configuration changes are available to assist in creating the images (example role types: login, node, storage, etc).

Images from a running system

If you have nodes already running a supported operating system, or prefer to setup a node by booting the installation media and going through the distro supplied installer that is supported as well. Once the node has been configured and the CluserVisor client package has been installed, you can take an image from the working system. This will copy the entire operating system and optionally disk configuration to the ClusterVisor appliance or server. Once the image has been captured it can be deployed on other nodes just like an image from a Linux distribution.

Images can be updated from a running system at any time, or new images can be created. This is a great way to try operating system updates or changes, and be able to roll-back if the desired effect isn’t what you wanted.

Disk Layouts

When deploying stateful images to nodes ClusterVisor abstracts the image from the “disk layout” The disk layout contains information about partitioning, software RAIDs (via MD), LVMs, bootloaders, filesystems, and mount points. This allows you to reuse one image across multiple machines even if one node might have two NVMe drives in a software RAID1 vs another machine that just might have one SATA drive.

Deploying images to nodes

ClusterVisor manages the entire boot process for nodes, and at any point you can set a node to be re-imaged from the ClusterVisor server on next boot. Images can be deployed unicast to a single node at a time, or multicast if installing lots of nodes in the cluster. You can configure a unique image and disk layout to each node, and during imaging it will re-partition, format and do all other setup necessary on the disks, copy the image from the ClusterVisor server to those disks, and then apply any node specific configuration from the ClusterVisor plugins (i.e. networking, nfs mounts, etc – see Configuration Management section). If you are only deploying a few nodes you can use the default unicast method, or if you are deploying lots of nodes images can be multicast from the ClusterVisor appliance or server to reduce bandwidth and make installing 10s to 100s of nodes very quick.

Stateless images

Any image can be deployed as stateless inside of ClusterVisor. These images are downloaded from the ClusterVisor appliance or server on a node boot. The entire image is stored in RAM on the node and will always return to a known state on reboot. To support lots of nodes stateless booting at once, multiple stateless image servers besides the ClusterVisor appliance can be dedicated, and then nodes will randomly select from one of the stateless servers on boot.

Stateless can be a great solution for high security environments, since the machines will be wiped on every power cycle.

Request a Consultation from our team of HPC and AI Experts

Would you like to speak to one of our HPC or AI experts? We are here to help you. Submit your details, and we'll be in touch shortly.

  • This field is for validation purposes and should be left unchanged.