Technical workshop: High-Availability Groups

The OrionVM platform portal was designed to make provisioning cloud infrastructure as simple as possible by mimicking a traditional hypervisor environment, in lieu of the commonplace cornucopia of complicated, confusing cloud computing. If only OrionVM paid me extra for alliteration.

Today we’re exploring node isolation and High-Availability Groups, an example of a platform feature that helps teams provision cloud infrastructure with additional confidence.

High Availability (HA) is a design practice that permits an environment to continue operating if one or more components are offline, not accepting requests, or are operating in a degraded state. This can be achieved by strategic use of redundancy, and careful design of applications, platforms, and networks.

OrionVM does this automatically and transparently for storage. The platform’s low-latency InfiniBand stack exports all storage in a data centre to a virtual SAN (vSAN), with virtual disks synchronously replicated across multiple redundant nodes for minimum n+1 redundancy. This occurs regardless of the performance class of the virtual disk, whether it be NVMe, enterprise SSD, or archival spinning. In the event of storage or a node going offline, IO requests are handled by other replicas in the cluster automatically, while redundancy is restored onto other hardware.

High-Availability Groups provide a similar isolation mechanism for virtual compute. Virtual servers grouped for HA are guaranteed not to run on the same hypervisors. They operate by informing the platform orchestrator in each data centre to override its standard allocation system and isolate virtual servers if it otherwise would have placed them on the same node.

This can be useful for grouping the following types of virtual servers where a failure of one can be mitigated by operating another:

  • Web application servers
  • Load balancers
  • Database servers, with one or more read replicas or standby servers
  • Application hosting for Windows Server

This mechanism can be deployed in two ways. The first is to create a new set of VMs off one you are in the process of provisioning. This can be done from the New Instance screen:

  1. Log into the End-User Portal and choose Instances
  2. Click New Instance, and click Advanced
  3. From HA Duplicates, select One or Two

Another approach is to take an existing virtual server fleet and group them for HA:

  1. Log into the End-User Portal and choose Instances
  2. Shift-click one or more virtual servers you wish to group
  3. Right-click and choose Group for High Availability

High Availability Groups enforce node isolation within the same data centre upon boot. Thus, all but one virtual server needs to be offline to group them in this fashion.

In the next workshop, we’ll take a look at how this works in practice at an application level, and how you can use OrionVM’s Shared IP features to configure a basic heartbeat system.

Contact your account manager or our sales team if you’d like a deeper dive into how our platform features work, advice on best practices, or how you can build or migrate a solution on us.