restart: Unsupported value: "OnFailure": supported values: Always

[surajssd]

Using this docker-compose file

$ cat docker-compose.yml 
version: "2"
services:
  mariadb:
    image: centos/mariadb
    restart: on-failure

which fails as

$ kompose up
WARN[0000] [mariadb] Service cannot be created because of missing port. 
We are going to create Kubernetes Deployments, Services and PersistentVolumeClaims for your Dockerized application. 
If you need different kind of resources, use the 'kompose convert' and 'kubectl create -f' commands instead. 

FATA[0000] Error while deploying application: Deployment.extensions "mariadb" is invalid: spec.template.spec.restartPolicy: Unsupported value: "OnFailure": supported values: Always 

when converted it looks like this:

$ cat mariadb.yml 
apiVersion: extensions/v1beta1
kind: Deployment
metadata:
  creationTimestamp: null
  name: mariadb
spec:
  replicas: 1
  strategy: {}
  template:
    metadata:
      creationTimestamp: null
      labels:
        service: mariadb
    spec:
      containers:
      - image: centos/mariadb
        name: mariadb
        resources: {}
      restartPolicy: OnFailure
status: {}

so this cannot be OnFailure it always has to be Always which stops us from using docker-compose's restart: on-failure to simulate kubernetes jobs.

[surajssd]

So I tried this in continuation with discussion at #236 (comment)

[surajssd]

But this is perfectly valid thing

$ cat pod-maria.yml 
apiVersion: v1
kind: Pod
metadata:
  creationTimestamp: null
  name: mariadb
spec:
  containers:
  - image: centos/mariadb
    name: mariadb
  restartPolicy: OnFailure

a pod without controller can have restartPolicy as OnFailure

[kadel]

I'm afraid that this doesn't have easy solution.
I don't think that we can create just pod if there is restartPolicy: on-failure in docker-file.

But we could get away with just ignoring it, and showing warning explaining that this is ignored and Always will be used instead)

[surajssd]

But we could get away with just ignoring it, and showing warning explaining that this is ignored and Always will be used instead)

@kadel but then this is not solving our original problem of how do we simulate k8s-job like workflow with kompose and docker-compose?

[kadel]

I'm not even sure if we should simulate Kubernetes job workflow with kompose and docker-compose.

You are proposing creating Job if there is restart: on-failure?

[ngtuna]

so with #334 pod will be created instead of Job if there is restart: on-failure.

[surajssd]

@ngtuna yes you are right!

[surajssd]

With #334 merged this issue is solved.

[mhyousefi]

So, what is the point of having deployment.spec.template.spec.restartPolicy if Always is only allowed option?????

Also, I suppose one could use backoffLimit: 0, though it doesn't seem like a robust solution as the reason why the pod terminated could very well be an actual failure!