Chapter 2. RHOSP server group configuration for HA instances
Create an instance server group before you create the RHOSP HA cluster node instances. Group the instances by affinity policy. If you configure multiple clusters, ensure that you have only one server group per cluster.
The affinity policy you set for the server group can determine whether the cluster remains operational if the hypervisor fails.
The default affinity policy is affinity. With this affinity policy, all of the cluster nodes could be created on the same RHOSP hypervisor. In this case, if the hypervisor fails the entire cluster fails. For this reason, set an affinity policy for the server group of anti-affinity or soft-anti-affinity.
-
With an affinity policy of
anti-affinity, the server group allows only one cluster node per Compute node. Attempting to create more cluster nodes than Compute nodes generates an error. While this configuration provides the highest protection against RHOSP hypervisor failures, it may require more resources to deploy large clusters than you have available. -
With an affinity policy of
soft-anti-affinity, the server group distributes cluster nodes as evenly as possible across all Compute nodes. Although this provides less protection against hypervisor failures than a policy ofanti-affinity, it provides a greater level of high availability than an affinity policy ofaffinity.
Determining the server group affinity policy for your deployment requires balancing your cluster needs against the resources you have available by taking the following cluster components into account:
- The number of nodes in the cluster
- The number of RHOSP Compute nodes available
- The number of nodes required for cluster quorum to retain cluster operations
For information about affinity and creating an instance server group, Compute scheduler filters and the Command Line Interface Reference.
