This message was deleted.

Interesting. So I did two things which seemed like they fixed it: 1. Instead of

image.imageUri

, I changed it to

image.imageName

- but that wasn’t sufficient. It just took too long and the call failed because of a 20 minute timeout (

timeout while waiting for state to become 'tfSTABLE' (last state: 'tfPENDING', timeout: 20m0s)

) 2. I also had to run it one time with the following environmental variable:

PULUMI_DEBUG_PROMISE_LEAKS=true

and then it ran to completion 🤔

Yes, imageName is the correct property of Image to use for that property. The other part is interesting. At a guess, I'd say you could have achieved the same thing by deleting the earlier version of the task definition via the ECS console. It looks like it was related to something being out of sync in AWS.