Kestra 2.0 Released: New Engine & Workers Anywhere

Ever since its initial release in February 2022, the open-source orchestration platform Kestra has evolved by listening closely to real-world deployment feedback. While the early version relied on a heavy combination of Kafka and Elasticsearch, the project quickly adapted by introducing a JDBC backend just four months later to simplify deployments.

Building upon years of operational insights, Kestra 2.0 officially shipped this month with a major rewrite of its core engine. In previous 1.x versions, every worker required a direct connection to the central database, creating severe deployment hurdles for security teams managing restricted enterprise networks.

Stock photo for illustration only, not from the actual event

The new 2.0 release resolves this by cleanly separating the platform into a control plane and a data plane. The control plane houses the executor, scheduler, webserver, indexer, and a new worker controller—none of which execute user code. Meanwhile, workers operate exclusively within the data plane where tasks are safely processed.

Workers now maintain a single persistent gRPC stream initiated entirely from the worker side. Because no inbound connections are required and workers hold no database credentials, they can securely execute tasks across different clouds, separate regions, or inside strict outbound-only on-premise networks.