Por nu 
Fspecifically with the resource type "file"
Rwhat s the best way to debug a flow / script stuck in a deployment state?
Rlook for the associated dependency job
T@rubenf I'm seeing the following migration error when upgrading from v1.554.0 to v1.561.0
Checking my db I don't see any null settings values
Which points me to the migration might be the thing that is trying to insert a null value: https://github.com/windmill-labs/windmill/blob/6dfa4ff5ba26bc4912b1be1a9709728cf6c2d119/backend/migrations/20251006143822_ducklake_instance_settings_better.down.sql#L3. I don't have ducklake settings
Checking my db I don't see any null settings values
Which points me to the migration might be the thing that is trying to insert a null value: https://github.com/windmill-labs/windmill/blob/6dfa4ff5ba26bc4912b1be1a9709728cf6c2d119/backend/migrations/20251006143822_ducklake_instance_settings_better.down.sql#L3. I don't have ducklake settings
R@Diego looks to be your migration:
Dhm
DCan you try to run this migration by hand
https://github.com/windmill-labs/windmill/blob/6dfa4ff5ba26bc4912b1be1a9709728cf6c2d119/backend/migrations/20250731132157_ducklake_instance_settings.up.sql
https://github.com/windmill-labs/windmill/blob/6dfa4ff5ba26bc4912b1be1a9709728cf6c2d119/backend/migrations/20250731132157_ducklake_instance_settings.up.sql
GitHub
Open-source developer platform to power your entire infra and turn scripts into webhooks, workflows and UIs. Fastest workflow engine (13x vs Airflow). Open-source alternative to Retool and Temporal...
DSomehow ducklake_user_pg_pwd does not exist
Tokay that fixed the issue, I'm wondering how my windmill db got in that state though since I never run any manual migrations. Did I miss an upgrade message about needing to run a manual migration?
Dno this should have been run automatically liked other migrations
Di don't understand either
D@rubenf if you have an idea how it's possible that a migration got skipped
RNot possible
Dwell it's magic 
Dnever happened before
Rsorry im not seeing any issues with depedency jobs eitehr
PIs it just having infinite spinner e.g. deployment in progress? Are lockfiles generated?
If there is no locks there is likely a djob somewhere still running.
Did you try to redeploy?
If there is no locks there is likely a djob somewhere still running.
Did you try to redeploy?
RI actually wiped my entire windmill instance and did a fresh push
R" Deployment in progress" on the flow page
PWhat version are you on? The minimal reproduction is to just create a flow and deploy it?
R1.555.1
PAny specific actions? Imports?
PAnd what is your infra? k8s? Docker? Native?
DDid you ever do something to your global settings that might have deleted it ?
DI'm going through the code and assuming the migration ran, I don't see how this was possible
Rright now a local docker running
Rits a rather complicated flow, all python is checked and valid with tests and i can confirm they work.
RI can view all scripts etc on the frontend
TI've been setting instance-level settings using a
instance_settings.yamloh
Dthere we go 
thanks
thanksT@Diego just to be clear I do feel like the way this has broken is not an optimal customer/developer experience. I don't think upgrading versions should break in this way.
Dyes it's a bug
PCould it just be locking things? Could happen if you have poor connectivity.
But if it is complicated flow, I think it is the best to gradually reduce until it works
But if it is complicated flow, I think it is the best to gradually reduce until it works
T@Diego gotcha! happy to be a point of contact on the fix if that's helpful 
really appreciate how you + ruben have hopped in here so quickly
really appreciate how you + ruben have hopped in here so quicklyRill refactor it and start with each step until it breaks somewhere
PYeah. But you should really check dependency jobs. You can filter by that job kind in run page
PI just think you have something heavy or it is somehow stuck on a a dependency
Rits a local instance, no jobs running except the deployment
Pby heavy I mean you have hundreds of dependencies (including transitive) and your instance didn't finish fetching all necessarily metadata
Rthey all finished, is there a chance one may have not?
PIf it is a single flow and there is no relative imports, it would be a single dependency job. If that dependency job has finished then the flow should be viewable as deployed
MThis is how simple it could be to interact with wmill storage / ducklake
MBun
Bun 1.3 introduces zero-config frontend development, unified SQL API, built-in Redis client, security enhancements, package catalogs, async stack traces, VS Code test integration, and Node.js compatibility improvements.
MVery impressive from bun, when I was reading it I kept thinking that these can be made to interoperate with windmill by default
MResources ≈ redis KV (types can be specified inline, defaulting to ‘state’ _resource_type: ‘s3’)
Variables ≈ secrets
Storage ≈ s3
* Ducklake ≈ sql
Variables ≈ secrets
Storage ≈ s3
* Ducklake ≈ sql
RAgreed @metapontum , we will probably want a simplified syntax, maybe a one function call in our sdk to setup the s3 client
ZHave you thought about Bun to build/serve the raw apps, @rubenf? (I’ve actually been using raw apps to iframe to a subdomain with a cookie.) Was reading through some of the raw apps branch and was just curious?
R@Zach raw apps are purely static frontend rendering + windmill jobs for background runnables