Chapter 1. Release notes

OpenShift Serverless Logic introduces a new data index mutation named ExecuteAfter, which enables you to create and execute a new workflow instance that can reuse the output of a previously completed workflow as its input.

The ExecuteAfter mutation accepts the following arguments:

Before this update, the KafkaSource dispatcher stopped committing offsets when the offsets of produced events were not consecutive integers, for example, when events were produced within a Kafka transaction. This behavior caused the dispatcher to stall and prevented subsequent events from being processed.

With this update, the KafkaSource dispatcher has been fixed to handle such empty offsets correctly. Additionally, the default Kafka consumer configuration for KafkaSource has been updated to isolation.level=read_committed. When Kafka transactions are used to produce events into a Kafka topic, the KafkaSource now processes only the events from committed transactions.

In some cases, cluster-scoped resources such as webhook configurations are not removed during the uninstallation, reinstallation, or upgrade of the KnativeServing or Serverless Operator components. When this occurs, the reconciliation of KnativeServing fails, and the installation process becomes stuck with an error similar to the following example:

failed to apply non rbac manifest: Internal error occurred: failed calling webhook "webhook.serving.knative.dev": failed to call webhook: Post "https://webhook.knative-serving.svc:443/?timeout=10s": no endpoints available for service "webhook"

Some operations of the OpenShift Serverless Function MCP server, such as build and deploy, fail when triggered from the Cursor IDE using its built-in agent. When invoking these operations, the Cursor agent sends a malformed request for any optional parameters. Although the parameter values appear correctly formatted, for example, "quay.io/myuser", the OpenShift Serverless Function MCP API returns the following error message:

Error calling tool: Parameter 'optionalStr' must be of type null,string, got string

As of OpenShift Serverless 1.37 release, the kn client is built with RHEL 9 dependencies and cannot run on RHEL 8. Attempting to run the binary on RHEL 8 displays an error similar to the following:

kn: /lib64/libc.so.6: version `GLIBC_2.33' not found (required by kn)

OpenShift Serverless Eventing introduces EventTransform, a new API resource that you can use to declaratively transform JSON events without writing custom code. With EventTransform, you can modify attributes, extract or reshape data, and streamline event flows across systems. Common use cases include event enrichment, format conversion, and request-response transformation. EventTransform integrates seamlessly with Knative sources, triggers, and brokers, enhancing interoperability in event-driven architectures. This feature is now available as a Technology Preview.

See the following key features of EventTransform:

When starting a workflow via HTTP, you can now include additional properties alongside the workflowdata field in the request body. These extra fields are ignored by the runtime but are available in the Data Index as process variables as shown in the following example:

{"workflowdata": {"name": "John"}, "groupKey": "follower"}

You can now filter workflow instances by the content of workflow variables using GraphQL queries on ProcessInstances.variables. For example, the following query retrieves process instances where the language field in workflowdata equals Spanish:

ProcessInstances (where:{variables:{workflowdata:{language:{equal:Spanish}}}}) {
    variables,
    state,
    lastUpdate,
    nodes {
      name
    }
}

Previously, the Knative Kafka dispatcher could stop consuming events if a Kafka consumer group rebalance occurred while a sink was processing events out of order. This behavior triggered the following errors:

This issue is now fixed. The dispatcher correctly handles out-of-order event consumption during rebalancing and continues processing events without interruption.