Dagster - One Deployment Multiple Code Locations
ethanleifer opened this issue · comments
Ethan Leifer commented
Right now each chain deployment needs to each run its own dagster server which is specified through the workspace.yaml
in the dags package
This looks like this currently:
load_from:
- python_module: dags
execution:
config:
multiprocess:
start_method:
forkserver:
We can change it to the following
load_from:
- python_file:
relative_path: path/to/dataengineering_spark_team.py
location_name: dataengineering_spark_team_py_38_virtual_env
executable_path: venvs/path/to/dataengineering_spark_team/bin/python
- python_file:
relative_path: path/to/team_code_location.py
location_name: ml_team_py_36_virtual_env
executable_path: venvs/path/to/ml_tensorflow/bin/python
This will let us only have to deploy dagster once and we can just update the code locations. We can still keep the separate workspace.yaml for local development too.