This section provides a quick overview of some important concepts in Flyte with links to deeper dives.
Tasks are at the core of Flyte. A Task is any independent unit of processing. Tasks can be pure functions or functions with side-effects. Tasks also have configuration and requirements specification associated with each definition of the task.
Workflows are programs that are guranteed to reach a terminal state eventually. They are represented as Directed Acyclic Graphs (DAGs) expressed in protobuf. The Flyte specification language expresses DAGs with branches, parallel steps and nested Workflows. Workflow can optionally specify typed inputs and produce typed outputs, which are captured by the framework. Workflows are composed of one or more Nodes. A Node is an encapsulation of an instance of a Task.
Executions are instances of workflows, nodes or tasks created in the system as a result of a user-requested execution or a scheduled execution.
Projects are a multi-tenancy primitive in Flyte that allow logical grouping of Flyte workflows and tasks. Projects often correspond to source code repositories. For example the project Save Water may include multiple Workflows that analyze wastage of water etc.
Domains enable workflows to be executed in different environments, with separate resource isolation and feature configuration.
Launchplans provide a mechanism to specialize input parameters for workflows associated different schedules.
Registration is the process of uploading a workflow and its task definitions to the FlyteAdmin service. Registration creates an inventory of available tasks, workflows and launchplans declared per project and domain. A scheduled or on-demand execution can then be launched against one of the registered entities.
- Overview of the Execution of a Workflow
- Understanding Registration process