Skip to content

Latest commit

 

History

History
48 lines (42 loc) · 1.62 KB

design.md

File metadata and controls

48 lines (42 loc) · 1.62 KB

Design Example

OpenShift
- AZ0 default openstack-k8s-operators pods
- AZ1 pods for cinder-volume-az1-0 and glance-az1-edge-api-*
- AZ2 pods for cinder-volume-az2-0 and glance-az2-edge-api-*

AZ0: (ceph0)
- edpm-compute-0
- edpm-compute-1
- edpm-compute-2

AZ1: (ceph1)
- edpm-compute-3
- edpm-compute-4
- edpm-compute-5

AZ2: (ceph2)
- edpm-compute-6
- edpm-compute-7
- edpm-compute-8

AZ3: (ephemeral)
- edpm-compute-9
  • Nine compute nodes are deployed with 3 nodes per AZ; plus 1 with ephemeral storage.
  • Three three-node Ceph clusters are deployed per AZ (with the exception of AZ3)
  • The deployment of AZ0 resources will look very similar to the HCI example from the Development Preview.
  • AZ1, AZ2 and AZ3 are each managed as a separate OpenStackDataPlaneNodeSet.
  • AZ0 does not need to be physically in the same data center as OpenShift unless there high network latency between AZ0 and OpenShift.
  • The ceph cluster in AZ0 is the default storage backend for Glance, but the default Glance pod will be configured to access all Ceph clusters so that images may be copied between AZs.
  • Compute nodes in AZ-N will query the glance-N pod which will direct them to get them to use thier physically local Ceph cluster for all N.
  • The data path remains local though the control path is stretched across AZs.
  • The pattern can continue for more AZs
  • Node Selectors can be used to run replicas for glance on separate worker nodes