Hello there! Does anyone know of a workaround for https://github.com/pulumi/pulumi-kubernetes/issues/609 (Add await logic for DaemonSets)? My service is deployed as a DaemonSet and we’ve been running into some stability issues in our deployment pipeline lately. Currently we’re alerting with kube-state-metrics to detect the problem, but it’d be way nicer if

pulumi up

failed when some or all pods from our DaemonSet could not become ready within some duration. I’ve been crawling through other issues to try and find solutions and found a few that are relevant: • https://github.com/pulumi/pulumi-kubernetes/issues/1260 (Support user-specified readiness/await logic) • https://github.com/pulumi/pulumi-kubernetes/issues/1056 (CustomResource lifecycle causes issues with importing created resources) • https://github.com/pulumi/pulumi/issues/1691 (Support custom logic in resource lifecycle) And through those I’ve found a couple of promising pointers: • https://gist.github.com/timmyers/4d2fed53a358d4c98557a5886ae2afbb • https://github.com/pulumi/pulumi/issues/1691#issuecomment-573823609 But I can’t quite piece together how I’d actually block

pulumi up

on the DaemonSet becoming ready and throw an error. So far all I’ve been able to do is get the DaemonSet status showing as a stack output, but that’s not too helpful on its own 😅

This is veery hacky but you could use the

Command

resource from

@pulumi/command

to run the

kubectl

CLI in a loop forever until the daemonset pods are ready, and set a customTimeout for the Command resource

Oh yeah I could even use something like

kubectl rollout status

to wait for the DaemonSet deploy to complete, right? I’ll have to check, but I don’t think I have a

kubectl

binary (or a corresponding config) to use in my deployment pipeline. Sounds like I’d have to at least get the

kubectl

binary. But maybe I could effectively get a

.kube/config

out of pulumi since it already has the necessary creds?

Yeah you’ll need the binary unless you opt for making the API call through something like curl.

Makes sense, thank you!