// the find
apache/dolphinscheduler
Apache DolphinScheduler is the modern data orchestration platform. Agile to create high performance workflow with low-code
Apache DolphinScheduler is a workflow orchestration platform aimed at data engineering teams who need to schedule and manage complex DAG-based pipelines across heterogeneous data infrastructure. It competes directly with Airflow and Azkaban, positioning itself as the low-code alternative with a drag-and-drop UI. Best fit for teams already deep in the Java/Hadoop/Spark ecosystem.
Decentralized multi-master/multi-worker architecture is a real differentiator — no single scheduler SPOF like Airflow's scheduler process. The task type library is genuinely wide: Flink, Spark, Hive, DataX, SageMaker, Kubernetes jobs, gRPC, SQL stored procedures — all first-class rather than bolted on via plugins. Workflow versioning on both definitions and instances is useful in production where you need to rerun last week's workflow version without touching the current one. The Python SDK (pydolphinscheduler) lets you define workflows as code rather than clicking through the UI, which makes it actually tractable for teams that want Git-managed pipeline definitions.
The project is heavily Chinese-origin with bilingual docs, and the English documentation lags noticeably behind the Chinese version — some task type pages are sparse or link to outdated 3.x content when the codebase is further along. Observability is weak out of the box: the built-in monitor shows node health but there's no native integration with OpenTelemetry or structured trace export; you're writing logs to files and hoping your ELK stack picks them up. The SPI plugin system is powerful but the documentation on writing custom task types or registry plugins is thin — you're reading source code to figure out the contracts. Deployment complexity is real: production requires coordinating masters, workers, alert server, API server, and a registry (ZooKeeper or etcd by default) — the standalone mode is fine for demos but the gap to a production cluster is steep.