Skip to lesson
Exit
Orchestration with Dagster1 / 2

1 min lesson

Dagster vs the field

Work through the cases in "Dagster vs the field", pairing each signal with the move that fits.

Step 1 of 2

Dagster vs the fieldbe ready to defend a choice, not recite features

Airflow

Ubiquitous, huge connector ecosystem

Task-centric: order, not data

Lineage and data-awareness are add-ons

Dagster

Asset-centric, lineage + staleness built in

Strong local dev, typing, partitions

Control plane that delegates compute out

Databricks Workflows

Native to the lakehouse, zero extra infra

Weaker cross-tool DAG and dev ergonomics

Locks orchestration inside one vendor

Interview move

The JD names Dagster as a strong plus and cites scaling Dagster as an example project, so expect a direct “why Dagster over Airflow?” Answer with the asset model and a trade-off, not a feature list: “Airflow is fine if all I need is ordered tasks, but on a lakehouse I want lineage and staleness as first-class properties and I want one graph spanning Spark, dbt and connectors. Dagster gives me that. The cost is a smaller ecosystem and a team that has to learn the asset mental model.”

Watch out

Do not pitch Dagster as a Spark replacement. It is a control plane, not a compute engine. If you imply Dagster will crunch billions of rows itself you will lose the room - the heavy lifting belongs on Databricks and Dagster orchestrates it.