High Availability

In order to deploy Zowe in high availability (HA) mode, it is necessary to set up a Parallel Sysplex® environment. A Parallel Sysplex is a cluster of z/OS® systems that cooperatively use certain hardware and software components to achieve a high-availability workload processing environment. A production instance with this High Availability setup is required to achieve the necessary availability.

Sysplex architecture and configuration

A Sysplex is required to make sure multiple Zowe instances can work together. For more configuration details, see Configuring Sysplex for high availability.

To enable high availability when Zowe runs in a Sysplex, it is necessary to meet the following requirements:

• The Zowe instance is installed on every LPAR.
• The API services are registered to each Zowe instance.
• A shared file system is created between LPARs in the Sysplex. For details, see How to share file systems in a Sysplex.
• z/OSMF High Availability mode is configured. For details, see Configuring z/OSMF high availability in Sysplex.
• The instance on every LPAR is started.

Configuration with high availability

The configuration for the specific instance is composed of the defaults in the main section and the overrides in the haInstances section of the zowe.yaml configuration file.

In this section, ha-instance represents any Zowe high availability instance ID. Every instance has an internal id and a section with overrides compared to the main configuration in the beginning of the zowe.yaml file. For more information, see Zowe YAML configuration reference.

Caching service setup and configuration

Zowe uses the Caching Service to centralize the state data persistent in high availability (HA) mode. This service can be used to share information between services.

If you are running the Caching Service on z/OS, there are three storage methods with their own characteristics:

• Infinispan (recommended)
  • Part of the Caching service
  • Does not need separate processes
  • Highly performant
• VSAM (deprecated)
  • Familiar to z/OS engineers
  • Slow
• Redis
  • Needs to run in Distributed world separately
  • Good for Kubernetes deployment