Planning Considerations

Preparing a Red Hat OpenShift Container Platform

When planning a Red Hat OpenShift Container Platform (RHOCP) installation, make sure to consider the storage connection to the IBM Storage Scale storage cluster. Several factors influence the architecture and the installation of a software-defined storage cluster for RHOCP.

Therefore, consider the following factors:

• The estimated size of the shared data pools across all architectures and clouds.
• The size of the RHOCP cluster and number of nodes.
• The cluster topology, including high-availability aspects in one or more physical machines and their distance or even across multiple data centers.
• The network topology for the RHOCP cluster and the access to the IBM Storage Scale cluster because of the high dependency on the network interaction.
• The IBM Storage Scale cluster option of integration with an existing IBM Storage Scale environment or new IBM Storage Scale cluster.

Dependent on the use case and the workload you run, you must plan the following aspects for the IBM Storage Scale storage cluster and the IBM Storage Scale Container Native components:

• CPU
• Memory
• Network
• Disk types (ECKD, SCSI or mixed)
• Installation and setup of IBM Storage Scale storage cluster
• Installation of IBM Storage Scale Container Native and the associated IBM Storage Scale Container Storage Interface (CSI) driver for the RHOCP cluster

See the Red Hat OpenShift documentation Installing on IBM Z and IBM LinuxONE for more information.

Preparing an IBM Storage Scale cluster

If you are planning to build a new IBM Storage Scale storage cluster, it is recommended to implement it on IBM Z or IBM LinuxONE in separate LPARs from the RHOCP cluster. For a high-availability solution, consider deploying a stretched IBM Storage Scale cluster across physical entities. It is advisable to set up at least 2-3 virtual IBM Storage Scale storage servers as guests in a hypervisor on IBM Z or IBM LinuxONE. Detailed documentation for installing an IBM Storage Scale cluster can be found at Getting started with IBM Storage Scale for IBM Z .

By establishing a network connection to the IBM Storage Scale cluster, you can implement IBM Storage Scale Container Native in your Red Hat OpenShift Container Platform cluster environment.

Reference architecture for IBM Storage Scale Container Native

IBM Storage Scale is a clustered file system that provides concurrent access to a single file system or set of file systems from multiple nodes. The nodes can be SAN attached, network that is attached, a mixture of SAN attached, and network attached, or in a shared-nothing cluster configuration.

IBM Storage Scale Container Native allows the deployment of a clustered file system in an RHOCP cluster. Using a remote mount-attached file system, the IBM Storage Scale solution provides a persistent data store to be accessed by the applications via the IBM Storage Scale Container Storage Interface (CSI) driver using Persistent Volumes (PVs).

Networking considerations

The network topology for an RHOCP environment is crucial because it influences the performance of the data access from the RHOCP applications, as well as the disk operations operated via CNSA to the IBM Storage Scale cluster.

The cross-cluster mount topology, which is like the one described in the NSD client/server model is highly dependent on the speed of the network to the IBM Storage Scale storage cluster.

With the IBM Z and IBM LinuxONE internal network capabilities like HiperSockets or shared network cards, you can accelerate the data access. At the same time, use the high security and crypto capabilities on the mainframe for all data access and at rest operations.

See Installing on IBM Z and IBM LinuxONE for more information.

Physical storage considerations

For the completeness of the storage decisions, it is important to mention that the storage that is certified to be attached to an IBM Z and IBM LinuxONE server can be used for an IBM Storage Scale storage cluster build on the mainframe. Therefore, you can use direct-attached ECKD type or FCP/SCSI type of storage servers. A preconfigured IBM ESS (Enterprise Spectrum Scale) storage server, or a storage server that is connected via IBM Spectrum Virtualize Family of Products (for example SAN Volume Controller – SVC) are additional options. The storage types can be mixed or especially used for tiering and you can define which data should be stored on which tier or storage class. That enables a high flexibility for a software-defined storage solution for RHOCP in conjunction with IBM Spectrum Scale. For details, how to create an IBM Storage Scale storage cluster, see the white paper Getting started with IBM Storage Scale for IBM Z.

High availability considerations

The implementation of an IBM Storage Scale storage cluster with more than two nodes is highly available per design. The implementation of the CNSA components in RHOCP as well.

The important part is the infrastructure on which the software components of the IBM Storage Scale server run. If the IBM Storage Scale storage servers are implemented on IBM Z, you can build an environment with high availability infrastructure, taking advantage of the mainframe characteristics. Depending on the metro distance (near distance) or global distance (far distance) the options are:

• A stretched cluster for IBM Storage Scale: Enables an active-active environment in metro distance with IBM Storage Scale Control, without physical disk replication requirements.
• A metro or global replicated setup for the IBM Storage Scale cluster: Enables an IBM Storage Scale controlled disk replication, without physical disk replication requirements.
• A high availability environment with IBM Z and replicated data via metro or global mirror: Requires a metro or global distance replication setup, which is controlled by the user. For example, by using disk replication functionality.
• A GDPS enabled HA / DR for an automated continuous operation environment: Enables an active monitored environment that uses GDPS in metro or global distance with physical disk replication, disk Hiperswap (ECKD disks) and application HA, with entire site failovers and recovery.

On IBM Z the use of three LPARs is a best practices setup that considers the HA requirements for RHOCP and can integrate with the HA requirements from IBM Storage Scale and CNSA.

• The RHOCP environment can be virtualized by using z/VM or RHEL KVM.
• Each LPAR hosts an RHOCP control plane and multiple RHOCP compute nodes with one IBM Storage Scale server node.
• In each RHOCP compute node or infrastructure node an instance of CNSA is run.
• Every IBM Storage Scale node has direct access to the storage disks through the FICON® or FCP/SCSI attachment

You can set up every IBM Storage Scale node with both direct access and access through network to another IBM Storage Scale node. If direct access is lost, the data access is assured through the other Scale nodes. The following picture illustrates an RHOCP cluster with CNSA components integrated in RHOCP in a cross-cluster mount model.

Resource considerations

The following section identifies the system-level requirements of all servers within your RHOCP environment and the IBM Storage Scale storage cluster. At the time of writing, the following software levels are recommended:

• OpenShift Container Platform version 4.14 or later
• IBM Storage Scale Container Native Storage Access 5.2.1.0
• IBM Storage Scale Container Storage Interface (CSI) 2.12.0
• IBM Storage Scale Data Access Edition or Data Management Edition 5.1.9.x for the storage cluster.

For more information about software levels and their compatibility, refer to Supported OS and software versions .

The network requirements are as follows:

• All nodes in the Red Hat OpenShift cluster should be able to communicate with all nodes of the IBM Storage Scale storage cluster.
• A minimum of 10 Gb Ethernet network is needed. However, dependent on your workload and the usage of your network, a 25 GB might be preferred.
• RDMA for InfiniBand® or RoCE for Ethernet is not supported.

The minimum resource requirements for a Red Hat OpenShift cluster installation are:

The hardware resources must be sized according to your workload.

The requirements for your IBM Storage Scale storage cluster highly depend on the workload you plan to run on this cluster.

When you plan for an IBM Storage Scale cluster, keep in mind that IBM Storage Scale itself needs a considerable amount of memory. Most of the Spectrum Scale memory usage is for the Spectrum Scale page pool, but that is not the only memory that is used by Spectrum Scale. The page pool value should be one third to one half of the memory on the node. As a result, the minimum memory size for a Linux instance should be 4 GB.

The actual memory that is used depends highly on the other workload running on a cluster node. When defining the Spectrum Scale page pool size, you must consider other functions that might be running on the node.

See Hardware requirements for more information about worker node and pod memory requirements.