Server groups

Server groups define collections of virtual machines (VMs) that can be given specific properties. For example, a server group policy can state that the VMs in the group must not be placed on the same physical hardware, for availability reasons.

Server groups are specific to each project and cannot be shared across projects.

There are four policy types to choose from for a server group:

• Affinity: defines a strict affinity policy. Under this rule the scheduler is required to run a new compute instance (VM) on the same hardware as another existing instance in the same server group. If that hardware does not have enough resources to run the new instance alongside an existing one from the same server group, the scheduler will not run it.

• Anti-affinity: defines a strict anti-affinity policy. Under this rule the scheduler is required to run a new compute instance (VM) on different hardware from another existing instance in the same server group. If no other hardware has enough resources to run the new compute instance (VM), the scheduler will not run it.

• Soft affinity: defines a flexible policy. Under this rule a new compute instance (VM) tries to run on the same hardware as another existing instance in the same server group. If that hardware does not have enough resources to run the new instance alongside an existing one from the same server group, the scheduler will try other hardware.

• Soft anti-affinity: defines a flexible anti-affinity policy. Under this rule the scheduler tries to run a new compute instance (VM) on different hardware from another existing instance in the same server group. If no other hardware has enough resources and the only capacity available is on the same hardware as an instance from the same server group, the scheduler will not run it.

Affinity policies are typically used by architectures that require guaranteed throughput, and anti-affinity policies by architectures that require availability.

Next steps

• Compute instances
• Planning your instances