What's new in Edge Flow Manager

Edge Flow Manager (EFM) enables you to manage, control, and monitor agents that are deployed on the edge devices in IoT implementations. You can use these agents to collect real-time data originating from the devices to create and push actionable intelligence and insights to the place of data origin.

Edge Flow Manager 1.5.0 is a feature release that contains new features, performance improvements, and bug fixes, including:
Manifest Refresh functionality

It is now possible to refresh the manifest behind an agent class. The manifest drives what kind of processors and services are available in the designer. Earlier if agents were upgraded to newer versions, existing flows in the flow designer did not reflect the new capabilities. With the manifest refresh functionality, you can refresh the manifest for existing flows. For more information, see Refreshing the agent manifest for existing flows.

Automatic Agent cleanup

It is now possible to automatically clean up stale agents. So if an agent is decommissioned, it will not stay in the agent list as an offline agent forever. You can configure the timeout value globally and on agent class level as well. If timeout is not configured, the original behavior of never cleaning up missing agents is applied.

Related properties:
  • efm.monitor.agentCleanupAfterInactivityDuration=1d
  • efm.monitor.agentCleanupAfterInactivityDurationForClass.[your\ class\ here]=6h
Pushing Assets to Agents – S3 integration

This functionality is now generally available and can be used in production environments.

You can push assets to the agents originated from an S3 bucket (previously the asset had to be manually placed on each EFM node), so the agent can download any arbitrary asset needed for a processor. For example, you can push enhanced Machine Learning models to support decision making on the edge. For more information, see Using Asset Push command.

Support for flow.json publishing

With this new functionality, EFM is can serialize flows in JSON format, while maintaining backward compatibility. EFM can still serialize flows to YAML format as well depending on the supported formats of the agent.

The following endpoint has changed: /flows/{id} – model class Flow now has a list of flow content objects instead of flow content / flow format field.

If you have existing flows, you need to republish them so that the flow will be transformed to JSON format as well on the EFM side. This way new agents joining existing agent classes will be able to download in JSON format too.
Removal of support for legacy MiNiFi Java releases

Cloudera no longer supports Legacy MiNiFi Java versions 0.6.0.1.3.1.0 or lower. If you use a legacy version, you should upgrade to version 1.22.07 or higher. For more information, see CEM component compatibility.

For more information on bug fixes, see Fixed issues.