Project Running Sequence (Magic xpi 4.5)

The following diagram describes the sequence of events that occur when your project's status is RUNNING:

The Magic xpi servers (engines) communicate with the Magic Space through a GigaSpaces proxy (the middleware layer) that connects client applications to the Magic Space. The Magic Space can reside on a number of processes and a number of machines as a single unit. It is the responsibility of the GigaSpaces proxy to make the right connections between the servers and the Space partitions.

1. When the project is running, three elements operate simultaneously: workers, polling triggers, and external triggers. Each Magic xpi server can run one or more workers/triggers for a variety of tasks:

   1. A polling trigger (such as the Directory Scanner), also known as an asynchronous trigger, is a Magic xpi server thread that constantly checks external systems (such as an email account) to see whether it needs to invoke a flow without waiting for a response. When it needs to invoke a flow, it places a flow message with a status of READY_FOR_USE in the Magic Space.

   2. Each available worker scans the Magic Space for jobs to execute, meaning messages whose status is READY_FOR_USE. When the worker finds a message, it changes the status to IN_PROCESS, passes the message payload to the flow, and executes the flow logic. (This is known as a pull mechanism.)

2. When the main flow is completed, the worker updates the request message’s status in the Magic Space to DONE, and is then free to scan for additional messages.

3. If the message is from a synchronic trigger (such as an HTTP trigger), when the flow ends, the worker writes a response message to the Magic Space, which the trigger will send to the client.

4. Every parallel and stand-alone branch is also handled as a separate message written by the flow to the Magic Space. This new thread (the parallel or stand-alone branch) can be handled by any of the project's workers, even a worker running on a different machine.