A "master" is a FlowForce Server instance that continuously evaluates job-triggering conditions and provides the FlowForce service interface. A master is aware of worker machines in the same cluster and may be configured to assign job instances to them, in addition to (or instead of) processing job instances itself.
Immediately after installation, the FlowForce Server instance acts as the master of a one-machine cluster (which includes itself). However, work will not yet be distributed, since there are no workers to take over the workload. To set up a cluster, install additional FlowForce Server instances and convert them to "worker" mode, as shown further in this documentation. A cluster ready for load balancing is assumed to be set up when at least one machine acts as worker, in addition to the master machine.
|Note:||Only one master machine can exist in a cluster; the number of workers is not limited.|
There is no difference between operating a standalone FlowForce Server instance compared to a master instance. You configure jobs and view the processing log in exactly the same way. The only difference is that a master communicates with workers from the same cluster. In the cluster management page, you can view at all times the list of workers joined to the master, including those that attempted to join but did not confirm the security token. From this page, you can generate security tokens to confirm workers as such, and you can also remove workers completely. For further information, see Converting FlowForce Server to "Worker" Mode and Removing a worker from the master.
The master machine is responsible for continuous provision of service, collecting the status of job instances assigned to workers, and reporting the outcome. For this reason, it is important that the master machine is balanced according to the demands of your processing environment. To achieve that, you can redirect some or all jobs into queues that will be processed by workers, while the master will mainly provide the service interface. The master may also be configured to take some processing workload itself, in the event that no workers are available, see Setting up Distributed Execution.