User and service requests are passed through a load balancer to the Pega Platform servers. The Pega Platform server in turn makes requests to a database.
All load balancers for both on-premises and cloud Pega Platform installations use a single quiesce implementation. This allows for a quiesced node to have all users (both new and existing) be passivated immediately and for the node to be removed simultaneously.
In a highly available production system, the load balancer must have the following minimum requirements:
- Session-based affinity support to ensure that requests from one user are directed to a specific application server that maintains the state for that user
- Ability to disable a Pega Platform server or take the Pega Platform server out of the load balancer rotation to facilitate shutdown
- Automatic active or passive Pega Platform server health monitoring for Pega Platform server failure detection
- Scripting to manage cookies, invalidation, and optionally for allocation and deallocation of Pega Platform servers based on production load