Skip to main content

Fluss

Fluss is the Arkitekt service for managing the design of workflows and their runs. It provides the storage backend for importing and exporting workflows, and acts as the log service for workflow runs.

info

Fluss is not a workflow engine. It does not execute workflows, but rather manages their data for design and runs. Workflow execution in an Arkitekt system is handled by Apps not Services. To read more about Scheduling Apps and their execution, see the Workflow documentation.

Open Fluss on Github

Fluss Design​

An Architectural Overview of the Fluss Service. And its relationship with the execution of workflows by an Arkitekt App.

The Workflow Format​

A Workflow that is designed by a user (in an App like Orkestrator) is stored in Fluss in a plain JSON format. This JSON format contains the definition of the worklflow by defining the nodes and their connections. While you can manipulate this JSON directly, the JSON format is not primarily designed for human readability. Instead, please use the Orkestrator App to design and manage workflows.

We also provide webcomponents that allow you to render the workflow in a visual format. This is useful for displaying the workflow in a user interface (just like here in the documentation).

Workflow Runs​

When a workflow is executed by an App, the App will create a new Run in Fluss. This run will contain the log of all the events that occur during the execution of the workflow. This log is stored in Fluss and can be queried by any app. While these logs provide a detailed view of the logical execution of the workflow, they do not contain all the details for provenance reasons. For example (quite obviously) the logs do not contain the actual data that was processed by the workflow. Instead, the logs contain references to the data, and the data is stored in the storage service. However they do contain the details of the processing steps that were applied to the data.

Because a normal workflow can be quite complex, the logs can be quite large. Fluss lets you filter the logs by time, by node, by status, and by other criteria, and with Orkestrator you can view the logs in a visual format.