Within each VERDE cluster, one server is singled out as a managing entity, functioning to control session traffic and distribute activity equally among available servers. This server is the Cluster Master.
The Cluster Master relies on VDI servers to provide and maintain the state information collected from the applications running on those systems. The Cluster Master utilizes this information to relay load balancing results in the form of referrals when new clients connect to the cluster. Existing clients are referred to the VDI servers already running active sessions.
There is only one active cluster master, but any number of VDI servers can be designated a Cluster Master candidate for fail over purposes. VERDE automatically assigns the role of cluster master to one of the candidate servers. A cluster master candidate is defined through the VERDE Menu.
More about the Cluster Master:
- Keeps track of:
- The status of the other nodes in the cluster.
- The License utilization, which is managed globally at the cluster level.
- Logged in users, overall and per server to balance the load across the entire cluster.
- A server designated as the Cluster Master can run with other VERDE Services, but it can also be configured as Cluster Master only. When a server is configured as a “Cluster Master Only,” it’s unnecessary to run the other VERDE Core services (such as Connection Broker, Hypervisor, CacheIO, SmartSync) that are usually required to run virtual desktops. By effect, the server specifications are much lower, allowing the server the ability to run a virtual machine if needed.
The file that acts as the functional core of a Cluster Master’s processes is located in central storage:
- Accessed through secure (https) port 8443. Any candidate must assign this port to Cluster Master communication.
- The Cluster Master is stateless; it maintains all state in RAM.
- When the Cluster Master node starts, its internal tables are empty.
- When VDI servers start, they immediately attempt to broadcast their system-level status (CPU load, total session count) to the Cluster Master, and retry every three seconds if previous attempts are unsuccessful.
- Practices a simple fail-over/recovery plan. If a Cluster Master fails, another candidate will be promoted to active Cluster Master, and VDI servers will continue to operate and retry to connect to the stand-in Cluster Master. Once a Cluster Master becomes available again, the existing state is automatically transmitted to it, and within seconds, the Cluster Master will contain all the information about the cluster that it missed when it was down.
- VDI servers broadcast session start/stop information to the cluster master as it happens, unless they are connecting (or reconnecting), in which case all the information is sent at once.
- The load balancing algorithm may result in new sessions always starting on the same server until that server's cumulative load rises to the level of the other servers.
- The VERDE Console – Management Console (MC) is also active on the same server as the Cluster Master. It works with the Cluster Master to fulfill user authentication and desktop provisioning, as well as VERDE Cluster Configuration and monitoring.