System Architecture
Power configuration
-
For HA operation, each server must be equipped with redundant power supplies. Each rack is configured as pairs of Power Distribution Units (PDUs). For consistency, connect all right-most PSUs to a right-side PDU and all left-most PSUs to a left-side PDU. Use as many PDUs as you need, in pairs. The following figure shows an example.
Figure 6. PSU to PDU power template
Cluster scaling guidance
Cluster scaling and sizing initiatives must account for the Openshift Container Platform cluster limits. For more information, see the Red Hat Cluster Limits documentation. The following table summarizes the limits:
Table 7. Openshift Container Platform cluster limits
Limit type
OpenShift v3.11 limit
Number of nodes
2,000
Number of pods
150,000
Number of pods per node
250
Number of pods per core
There is no default value. The maximum supported value is the number of pods per node.
Number of namespaces
10,000
Number of builds: Pipeline Strategy
10,000 (default pod RAM 512Mi)
Number of pods per namespace
3,000
Number of services
10,000
Number of services per namespace
5,000
Number of back-ends per service
5,000
Number of deployments per namespace
2,000
The guidelines in the following table apply to the deployment and scaling of Red Hat OpenShift Container Storage:
Table 8. OpenShift container storage limits
Limit type
OpenShift Container Storage v3.11 limit
Persistent volumes backed by the file interface
1,000
Persistent volumes backed by block-based storage
300
Storage cluster size
4 nodes at a minimum
When you design and specify the configuration of workload/application nodes, Dell recommends carefully sizing the cluster for a minimum number of nodes as a reference point for the total cluster cost. As a second data point relating to cost, you can size the nodes using the most cost-effective CPU and memory configuration. The optimal node configuration in many workload deployment sites lies somewhere between these options. As the cluster size and number of nodes increase, industry practice favors keeping the cluster size below four racks to avoid unnecessary complexity in network management and switch topology.
Even though deploying extremely large OpenShift container ecosystem clusters is possible, there are some disadvantages to managing such designs. In most container deployment sites, the largest cluster does not span more than two or three racks. As the need for container ecosystem clusters becomes established, it is often helpful to deploy two or more clusters so that workloads can be migrated from one cluster to another. An additional benefit of multiple clusters is the ability to release a cluster for reconfiguration or redeployment of the container infrastructure, although this is rare.