Dagster seems great - the testability angle is :ok...
# random
Dagster seems great - the testability angle is 👌 Guessing I'm just missing this somewhere in the documentation so apologizes for the silly question. AirFlow has the concept of a sensor (to wait for a file to arrive for instance). Is there a comparable feature for that or is the idea that a solid is so versatile you can implement using that Sorry for the newbie question, guessing other interested parties might have the same question! - and thanks! nicolas cage party dagsir
🎉 1
We currently have no explicit support for sensors. Broadly speaking, we think the Airflow approach (where a task is running and polling) is a misguided approach. However, users are free to reimplement that model in Dagster by having a polling loop within a solid. In our upcoming major release (exact date TBD), we will have support for sensors, although we will implement it in a different part of the system, within our dagster-native fault-tolerant scheduler process. “Sensing” will occur outside the context of a run, and when the appropriate condition arrives, a run will start. Hope that helps! And thanks for the kind words.
100% understand the sentiment - a busy sleep approach can have definite downsides on system load and even oddity in terms of maintenance (it would be great if these poking tasks never failed!) Would love to improve my own knowledge and thus of the teams. Any tips for a better way to approach such problems? Perhaps more trigger rather than busy wait? Again, thanks a lot! celebrate