It might be helpful to get more context from the user, are there cases where they couldn't just pull out the user/password/etc? Looks like it wouldn't be too hard postgresql://[user[:password]@][netloc][:port][/dbname][?param1=value1&...]

up

up

use current test in filename

@alangenfeld looks like only runs from user deployments have pipeline_snapshot_ids? Does debug export work without grpc servers?

Just realized this is a click feature, should we use this instead? https://click.palletsprojects.com/en/7.x/options/#values-from-environment-variables

feedback

Good point about the subchart. I don't love putting it under dagit- makes sense to us, but I don't think they seem related for users. Breaking change, but what if the subchart was under an additional yaml key? So that additional values related to user deployments could be under the userDeployments key?

I think the field would be more natural under userDeployments than dagit

Yay tests!

up

up

Any guess why it flakes on windows? Could be good to leave a note

1 fix, otherwise lgtm. Thank you!

Feedback

rebase

Missing links for run config and executor pages

links

Rendered, since tables are hard to read

Does that hold true if we use Instance.get()? https://dagster.phacility.com/D6883

user code containers should not need to redeploy if the Dagit instance changes.

This is ok with me, but we might benefit from another level of hierarchy under 'deployment'. It's a somewhat scattered collection of objects (coordinators/launchers/storages that are configured on the instance, executors in run config, and the daemon which is a service). However I do think all this stuff fits the bill of 'things you consider after you have written your first pipelines and are looking to host them somewhere'

Nice! Existing feedback looks good to me

I think this lgtm. Per offline discussion, might use something like kubeval rather than deserializing with the k8s client

How about The Queued run coordinator daemon has errors? Not running is an error, we surface it as one on the status page

Whoops! Thank you for fixing

feedback

There's an existing test that differentiates the all_daemons_healthy and all_daemons_live functions, but it doesn't actually invoke the cli. I guess it's time to bring the cli under test?

Looks good!