Just researching: - Does anyone have documentation...
# prefect-dbt
r
Just researching: • Does anyone have documentation or references on maxing out dbt dbt CORE capabilities before migrating to Prefect?
d
Hey Raul! I don't think I understand the question -- can you share a little more context?
r
Yup! So, I see a team in my company use Astronomer (Airflow) and dbt core. And they exp the following: • There are times where the processing on dbt jobs are segmented to unnecessary projects (fragile?) • Failed jobs block continuation of pipeline Has someone experienced something similar, documented their experience w/code examples, then migrated to Prefect?
d
Got it! Makes sense
We don't have a terribly complex dbt runner setup with Prefect (and we're still running Prefect 1) but some things that I did experiment with / have seen other folks do: • Split the dbt run into a "core" run for models that most/all downstream models depend on • Take a list of tags as an input for independent downstream runs for
run
,
test
,
snapshot
so that parts of a project can be run in isolation • Depending on which of those runs fail, reverse ETL & similar jobs may or may not be run depending on what you need
We've got a starting point 👉 https://github.com/PrefectHQ/prefect-dbt
Curious to see what other folks say!