dataflow/WARN/2026_001

Active Dataflow streaming jobs should have Streaming Engine enabled.

Product: Dataflow
Rule class: WARN - Something that is possibly wrong

Description

Streaming Engine is disabled for one or more active streaming Dataflow jobs.

Without Streaming Engine, worker VMs perform windowing, state management, and other resource-intensive operations locally. This legacy execution mode can lead to poor autoscaling performance, high VM resource utilization, and potential out-of-memory (OOM) issues.

Remediation

To enable Streaming Engine, configure the pipeline options based on the SDK language:

For Java SDK (version 2.10.0 or later)

Set the following pipeline option: --enableStreamingEngine

For Python SDK (version 2.21.0 or later)

Streaming Engine is enabled by default in newer Python SDK versions when using Python 3 and Runner v2. For older SDKs or to explicitly configure it:

  1. Enable Dataflow Runner v2 by setting the following option: --experiments=use_runner_v2 (Dataflow Runner v2 requires Streaming Engine for streaming jobs).
  2. Alternatively, explicitly enable it by setting: --enable_streaming_engine

Further information