I'm noticing significant slowdown for pulumi commands in TS projects utilising kube resources as the project expands. Running a trace shows that each new resource registration seems to incrementally add ~15s, with the vast majority of time being spent in pulumi-plan. This project only contains ~130 resources. Any obvious places I can look for what's causing this? edit: looks like this may be pulumi#14967. Disabling checkpoints improves performance dramatically

What are these resources (which type)?

They're a variety of different types, rbac rules, custom resources, deployments, etc. Crucially none of this time is actually spent deploying anything, I was only making modifications to 4 total resources during this run, but it took 30 minutes, 93% of which was spent in the 'pulumi-plan' stage on these registerResource events. I added another comment to the linked issue - I suspect this is due to the large states that pulumi-kubernetes ends up with (storing all the manifests in the state) and the amount of times it gets synced to the backend in full during a deployment

This is surprising to me for such a low number of resources. Thank you for sharing, we should take a closer look.

Does

ignoreChanges

cause Pulumi to not save the ignored properties in the state, or does it just ignore those parts during comparison? Curious if I can minimise the size of the state by ignoring things like

status

and other dynamic properties

It does save them to the state still

sad 😞

You could change

ignoreChanges

in your program and we'd still need compare the values, so the state needs them