What's New in Pentaho 11 for PDI Developers (and Should You Upgrade from 9.x or 10.x?)

Pentaho 11 is the new long-term support (LTS) release of Pentaho Data Integration and Analytics, following 10.2 from August 2024. Most of the headlines are about the server and the user console, so in this post I focus only on what matters to PDI developers: what changes in your day-to-day work, what might break, and whether you should upgrade from 9.x or 10.x now.

If you have not installed it yet, start with my PDI 11 installation guide.

1. Pipeline Designer: build transformations in the browser

The biggest visible change is Pipeline Designer, a browser-based editor for transformations and jobs. It works much like Spoon, but runs on a modern web UI, so you can build and edit pipelines without installing Spoon on every developer machine.

The important part for existing projects: it is compatible with the .ktr and .kjb files you already built in Spoon. Spoon is not going away, and you can mix the two. My take: keep using Spoon for heavy development and debugging for now, and try Pipeline Designer for quick edits and for team members who should not need a desktop install.

2. Projects: a real structure for your ETL

Before 11, PDI had no standard way to organize transformations, jobs and their configuration. Every team invented its own folder layout and kettle.properties conventions, and moving work between dev, test and production was often painful. Variables and file references could also resolve differently depending on how a job was launched.

Project Lifecycle Management adds projects as a first-class concept to fix this. Note that it is marked as a preview feature, so the behavior and UI may still change. I would try it on a new project first rather than restructuring a large existing codebase around it today.

3. Plugin Manager, and big data becomes a plugin

PDI 11 removes Karaf and OSGi completely. Big data components (Hadoop, Parquet, ORC, EMR and so on) are no longer a separate add-on. They are ordinary plugins, installed through the new Plugin Manager, and Pentaho says new features will ship as plugins wherever possible.

For developers this means:

• The PDI client and server are over 1 GB smaller.
• If a transformation opens with a "Missing plugins" error, the step you need is probably now a separate plugin. Install it from Plugin Manager.
• If you use custom or third-party plugins that were packaged as OSGi bundles (anything that lived under the Karaf system folder), test them before you upgrade and check with the author for an updated version.

4. Java 21

PDI 11 is certified on Java 21 (Oracle Java and OpenJDK), with Java 17 still supported. Java 21 means you can run an Oracle JVM without the old licensing costs, or any supported OpenJDK build such as Temurin or Azul Zulu.

If you are coming from 9.x on Java 8 or 11, this is the change most likely to trip you up. Custom Java code in User Defined Java Class steps and any extra JARs in your lib folder (JDBC drivers, utility libraries) should be tested on Java 21, since older libraries sometimes fail on newer Java versions.

5. OpenTelemetry observability

A new OTEL plugin sends PDI logs, traces and metrics in the OpenTelemetry format, which tools like Datadog, Splunk, Elastic, CloudWatch and Azure Monitor can read. You get:

• Logs collected in one place and shown in the job/transformation hierarchy
• Traces showing task timing, execution order and variable values
• Metrics to track data volumes at chosen points in a pipeline

If you have ever built your own logging tables and dashboards to find out which sub-job is slow, this is worth a look.

6. Easier Docker deployments

Pentaho now provides prebuilt, optimized Docker images for plain Docker, Kubernetes and the managed services (EKS, AKS, GKE), with standard paths and variables. Configuration files and licenses can be injected at startup. There are also images for running Carte, Pan and Kitchen in containers, which is handy if you want to run jobs as container tasks instead of on a long-lived server.

7. Bug fixes developers will notice

The 11.0 release notes list a long set of PDI fixes. A few that stand out for everyday development:

• Spoon running out of memory when running jobs in parallel, plus memory leaks when accessing repository jobs concurrently
• Performance problems with more than one Table Input in a transformation, and a reported slowdown in 9.3/10.2 compared with much older versions
• Main job not aborting after a sub-job fails
• PDI failing to start from a folder with spaces in the path
• Several steps not keeping variables in fields like FTP timeouts
• Kitchen sporadically failing with "Central Log Store is not initialized"
• Outdated SFTP library causing failures with newer OpenSSH servers

If any of these have been costing you workarounds, that alone may justify the upgrade.

8. Deprecated database dialects

Starting in 11.0, a number of older database types are deprecated in PDI, including Vertica, Hive (version 1), Ingres, Ingres VectorWise, InfiniDB and Neoview. They still work for now, but expect them to be removed in a future release. If your jobs depend on one of these, plan a migration or a generic JDBC connection.

Should you upgrade?

From 9.x: yes, plan it now

Pentaho 9.3 has been unsupported since July 1, 2026. That means no more fixes or security patches. Go straight to 11 rather than stopping at 10.2, since 11.0 is supported until at least July 2029 and you will only have to do the Java and plugin work once.

This is the bigger jump, mainly because of the move from Java 8/11 to Java 17/21 and the removal of OSGi. Budget time for testing, especially for custom Java code, extra JARs and big data jobs.

From 10.2: upgrade when it suits you

10.2 is also an LTS release, supported until at least March 2028. There is no urgency. Upgrade when one of these applies:

• You are hitting one of the bugs fixed in 11.0
• You want Java 21, Pipeline Designer, OTEL monitoring or the new Docker images
• You want to be on the release with the longest support runway

Since 10.2 already runs on Java 17, the jump is smaller. The main things to test are plugins affected by the OSGi removal and your big data jobs.

On the Developer Edition (former Community Edition)

Remember that the Developer Edition is licensed for development and non-production use, and new versions are now distributed through the academic program or trials. If your production jobs run on an old Community Edition, upgrading is also a licensing decision, not only a technical one. Check the terms before you plan anything.

Upgrade checklist

1. Install PDI 11 side by side in a new folder. Do not overwrite your current version.
2. Install Java 21 (or 17) and point PDI to it with PENTAHO_JAVA_HOME.
3. Back up your .kettle folder (kettle.properties, shared.xml, repository settings).
4. Copy your JDBC drivers and any extra JARs into the new lib folder.
5. Install the plugins you need (big data, Kafka, Salesforce and so on) through Plugin Manager.
6. Open every key transformation and job and look for "Missing plugins" errors.
7. Run your jobs with Kitchen and Pan against test data and compare row counts and outputs with the old version.
8. Check any jobs that use deprecated database types.
9. Only then switch schedules and production servers.

Wrapping up

PDI 11 is not a flashy release for Spoon users, but it is a meaningful one: a leaner runtime, modern Java, a browser-based designer, proper project structure and much better monitoring. For 9.x users the question is not whether to upgrade but how soon. For 10.2 users it is a good upgrade to schedule rather than rush.

References: What's New in 11.0, Release Notes 11.0, Components Reference, Pentaho Product Lifecycle.