OpenShift Container Platform for HA with OpenShift Container Storage
The accelerator is deployed as a minimal cluster of three worker nodes.
Each node supports one to three 500 GB disks for storage with the storage from each node replicated to two others within the system. Storage used by OpenShift Container Storage is obtained using a Persistent Volume Claim (PVC) and Persistent Volume (PV) from raw storage or local block storage.
Notes:
OpenShift Container Platform 4 - HA with OCSaccelerator is only supported on a single system and cloud groups that have a "Deployment type" of "Single Cloud" or "Multi Cloud".
| Number of storage nodes | Disks | Total capacity | Usable capacity | |
|---|---|---|---|---|
| Initial Configuration | 3 nodes | 1 disk of 500 GB on each node | 1.5 TB | 0.5 TB |
| 1 disk of 10 GB on each node |
- You can create application pods either on OpenShift Container Storage nodes or non-OpenShift Container Storage nodes and run your applications. However, it is recommended that you apply a taint to the nodes to mark them for exclusive OpenShift Container Storage use and not run your applications pods on these nodes. Because the tainted OpenShift nodes are dedicated to storage pods, they require only an OpenShift Container Storage subscription and not an OpenShift subscription.
- Each node on which OpenShift Container Storage runs requires 16 vCPUs and 64 GB of memory. These nodes are in addition to the worker nodes that are used to run application pods.
Table 2: OSD: Object-Based Storage Device
| vCPUs | Memory | Storage | Comments | |
|---|---|---|---|---|
| Starting node (OSD+MON) | 16 | 64 GB | 500 GB | 500 GB Storage + 10 GB per MON |
| OSD | 16 | 64 GB | 500 GB | 500 GB storage per disk (1 disk by default, scalable to 3 disks per node) |
OpenShift Container Storage requires large worker nodes to deploy. According to OpenShift Container Storage documentation, other than very specific services (e.g., monitoring and logging), only the OpenShift Container Storage pods must run on these nodes. It is accomplished by tagging the specific storage worker nodes for the OpenShift Container Storage pods to run on them, while including taints so that the application workloads can run on separate worker nodes. The result is that there may be a large number of VMs running in an Openshift Container Storage cluster, especially an HA cluster. The following nodes are available in a minimal HA cluster:
- 2 helper nodes
- 3 master nodes
- 2 worker nodes
- 3 storage nodes (specially tagged worker nodes)
- 1 bootstrap node
Note: Though the bootstrap node gets released after deployment, ensure you have the necessary resources for it. The minimum consumed amount of storage is 1.5 TB of which only 0.5 TB is available for workloads.