THIS IS A TEST INSTANCE. ALL YOUR CHANGES WILL BE LOST!!!!
...
- Core Framework Improvements
- Cluster Management logic has been stabilized and improved to better support zero-master clustering. For example, recovery time is faster now as we're not having to always wait some artificial length of time to see if any new nodes will join and thus we should reconsider which flow is the golden copy.
- The expression language capability now supports decimals, base64 and hex encoded values, and math functions.
- The flowfile, provenance, and content repositories have been reworked to support rollback. We've always been careful to ensure that upgrades would work well where existing flow state would be honored but what was harder was supporting the ability to rollback where state was written via a new version of NiFi but the user decided to rollback to an older version. This is now supported. This powerful feature also sets the stage for future work to enable rolling upgrades and automated rollbacks!
- Startup times for flows that have large backlogs should be far faster as the swap files have been reworked to provide summaries and avoid the need for full scans.
- For developers, they can now indicate that their processor should be given an instance isolated classloader. Some libraries, like Hadoop client and scripting engines as an example, use static variables that can pollute instances of processors on the graph. This feature allows those cases to be easily overcome by enabling isolation per instance of a processor. This also makes it really easy to expose classloader extension for custom jars to users.
- For developers, there is now an ability to migrate in-flight process session state to another process session. This will yield higher efficiency and makes for a far easier programming model for aggregation type patterns as seen in MergeContent, for example.
- User Experience Improvements
- We now provide visual indication of queue growth relative to back pressure settings and when back pressure is engaged. This will make the concept of congestion and back pressure far more intuitive and frankly it is just fun to see in the UI. Definitely check this out.
- After the 1.0.0 release several members of the community expressed how much they love the new look and feel but wished we had kept some of the colors. Better and more intuitive color contrast is back.
- Validation of components is now limited to occur only for components which are not scheduled to execute. This results in much faster UX behavior as many of the operations one could do through the UI and REST API would result in expensive validation operations that were unnecessary.
- Users can now export images of the provenance graphs.
- Users can now use cron-scheduling for components even on primary node only tasks.
- Multi-tenant authorization and internal authorization/policy management
- There is no longer just the limited Administrator, Data Flow Manager, etc. roles
- Policies are fully customizable for each component and at the controller level as well
- Updated Versions of Dependencies
- We now leverage the Azure Event Hubs 0.9.0 client library.
- We now interact with Apache Spark using the 2.0.1 libraries.
- We now interact with HDFS using the Apache Hadoop 2.7.3 libraries.
- New or Improved Processors
- New Fetch and Put processors to interact with ElasticSearch 5.0 and new processors to execute Query and Scroll operations against ElasticSearch.
- New processors to parse CEF formatted logs
- The Extract Email processors now support TNEF formatted attachments.
- New processor to validate CSV files.
- The Apache Solr processors have been updated to support SSL and Kerberos.
- New processors to act as client and server for Websockets.
- Templates export is now deterministic and thus can now be version controlled
A full list of issues that were resolved can be found at: https://issues.apache.org/jira/secure/ReleaseNote.jspa?projectId=12316020&version=12332640
...