This message was deleted.

I believe

pulumi refresh

will sync the currently understood state of the deployed stack resources to pulumi’s state.

pulumi up

will try to deploy a changeset using pulumi’s state. The deployment could fail if pulumi’s understanding of the stack is different than it actually is. Typically you don’t need to use

pulumi refresh

the times I’ve used it are when I made a manual change, or a change was secheduled, but hadn’t completed in the cloud yet, but the pulumi state thought it had.

you can add a

-r

but i also find that it will often want to update things that didnt change because the APIs version of it is sorted differently

Interesting! I didn't even see the

--refresh

flag. I think I'll add that to the CI then - there'd been a few times where I've been bitten by the expected state didn't match the actual. I'd rather have it doing extra updates than stopping on things that are incorrect.

I've been running

--refresh

for quite a while with

up

and it's never been a problem. I've found if you are using CloudFront it's required, because the ETag changes on each deploy.