Skaffold should strip or warn on imagePullPolicy: always when deploying to local cluster

[briandealwis]

@VeerMuchandi showed me a hard-to-diagnose situation where he was building and deploying to a local minikube cluster, but was encountering pull errors from the cluster:

Deploy completed in 1.032 second
Waiting for deployments to stabilize...
 - deployment/student-app-api: container student-app-api is waiting to start: spring-boot-student-app-api:fb068c00d2e1e08a81d57968d2b4defc228871d00115ba29374a027182abc9e4 can't be pulled
    - pod/student-app-api-65678c6ff6-blgmp: container student-app-api is waiting to start: spring-boot-student-app-api:fb068c00d2e1e08a81d57968d2b4defc228871d00115ba29374a027182abc9e4 can't be pulled
 - deployment/student-app-api failed. Error: container student-app-api is waiting to start: spring-boot-student-app-api:fb068c00d2e1e08a81d57968d2b4defc228871d00115ba29374a027182abc9e4 can't be pulled.

We verified that the images were present in Minikube's docker daemon.

It turns out the deployment had an imagePullPolicy: always.

Should we emit a warning when deploying a built image? Or perhaps this should cause an error?

[nkubala]

discussed this with the team. the consensus is that there's probably never a scenario where a user is trying to run skaffold with a manifest that has imagePullPolicy: always and doesn't wan't skaffold to succeed when deploying to a local cluster - otherwise they probably wouldn't be running skaffold in the first place 😄 it seems like ignoring the imagePullPolicy when deploying to a local cluster would be safe, and unblock the dev loop without users haven't to manually change their manifests.

[tejal29]

closed as duplicate of #6992