Tutorial pipeline works with dagit, errors out with dagster CLI

[gherka]

I'm slowly working through tutorial examples and stumbled across something odd - the iris_pipeline.py from this page works fine when I run it using dagit -f iris_pipeline.py, but throws up the following error when I run it using dagster pipeline execute -f iris_pipeline.py:

dagster.check.CheckError: Failure condition: Can not produce an instance reference for <dagster.core.instance.DagsterInstance object at 0x7fea42d3c290>

Is there something obvious that I'm missing?

[alangenfeld]

We should catch this with a better error message, but to explain whats happening

The environment DAGSTER_HOME controls where dagster will put things on the filesystem, such as marshaling data to the jupyter process for the notebook in the iris_pipeline. If this env var is not set, dagit will create a temporary directory to use that gets cleaned up when the dagit process exits. The dagster CLI on the other hand falls back to an in memory "instance", which fails as seen above when trying to prepare the jupyter subprocess.

You can read more about the "instance" here https://docs.dagster.io/overview/instances/dagster-instance#main

[alangenfeld]

error message improvements in https://dagster.phacility.com/D5158