I’ve been experimenting with the

aws-ts-hello-fargate

example on an M1 mac. When running

pulumi up

it never completes — the ECS service never gets healthy and is stuck in a death loop. This is due to building the docker container by default for arm64 arch instead of linux/amd64. I was able to modify the sample to specify the architecture in the docker build options and make progress. However, I’m concerned about the behavior when attempting to deploy a bad container. I had to break out out of

pulumi up

run

pulumi stack export

and clean up the ECS service from pending operations, then re-import using

pulumi stack import

. Even then when I ran

pulumi destroy

I had to manually cleanup the ECS cluster using the console. Is this expected behavior when a bad image is deployed?

Hmm this doesn’t sound optimal. Which example did you use?

hello pulumi ppl, i am a newb and I am experiencing a similar problem with my m1 mac mini. i have been working with one of the example projects in github (gcp-ts-k8s), and I can see that ‘pulumi up’ is clearly generating a ‘linux/arm64’ image instead of amd64. How can I explicitly specify the target arch? I couldn’t find the option in the “@pulumi/docker” api docs.