RTFM 🙂 that's one of the limitation, you've got to recreate the nodegroup the above is only true when switching from custom to not custom ami

i m basically only switching my nodegroup from one launch template id to another and that fails, doing the same with the cli :

aws eks update-nodegroup-version  --cluster-name ${CLUSTER} --nodegroup-name ${NODEGROUP_NAME} --launch-template name=${LT_NAME},version=${LT_VERSION}

so that starts to feel like a bug in the provider that's reusing the previous ouput of release version as an input even though in this case it should not

feels like a edge case because the resource is not aware of the content of the launch template, probably why this fails

pretty sure that's what happens, checking the state i see :

"outputs": {
    "__inputs": {
        "...": "...",
        "ciphertext": ....
    },
    "amiType": "CUSTOM",
    "arn": "....",
    "capacityType": "ON_DEMAND",
    "clusterName": "...",
    "id": "...",
    "instanceTypes": [],
    "labels": {
        "<http://node.kubernetes.io/role|node.kubernetes.io/role>": "workers"
    },
    "launchTemplate": {
        "id": "lt-03724fe1f358b6d39",
        "name": "node-template",
        "version": "1"
    },
    "nodeRole": "blahblah",
    "nodegroupName": "blahblah",
    "releaseVersion": "ami-05a6b4bea29fc7541",

so the

releaseVersion

got resolved to. the amiId (it does the same in the aws console)

only workaround i found was to perform a replacement when the lauchTemplate changes but it's not exactly cool if someone has any pointers to the code that handles that, not super familiar with the pulumi internals but got me curious 😊

created following issue https://github.com/pulumi/pulumi-aws-native/issues/1747