Deploying multiple container in kubernetes through a deployed app - kubernetes

Question

I am using minikube for deployment in my local machine.
I am deploying an app with help of helm charts. My deployment script looks like
apiVersion: apps/v1
kind: Deployment
metadata:
name: {{ template "app.fullname" . }}
spec:
replicas: {{ .Values.replicaCount }}
spec:
volumes:
- name: dockersock
hostPath:
path: "/var/run/docker.sock"
containers:
- name: {{ .Chart.Name }}
image: "{{ .Values.image.repository }}:{{ .Values.image.tag | default .Chart.AppVersion }}"
imagePullPolicy: {{ .Values.image.pullPolicy }}
volumeMounts:
- name: dockersock
mountPath: "/var/run/docker.sock"
command: ["/bin/sh", "-c"]
args:
- python3 runMyApp.py;
resources:
limits:
nvidia.com/gpu: {{ .Values.numGpus }}
My script runMyApp.py when executed launch 4 other containers and I want that Kubernetes deploy them as well at minikube and this is expected behavior from my understanding.
But When I other 4 containers they were deployed on host machine, just like I executed some docker run command on my host machine.
To verify that I was not mistaken,
so I tried to access them through other application using which is in minikube cluster but I couldn't. Then I tried accessing the application from the local environment and I was able to do that.
Is there any flaw in my understanding. If the behaviour is expected then what can I do to deploy other application in k8s as well.

Answer 1

In order to create new pods from within your python script, you should use the Kubernetes Python client library: https://github.com/kubernetes-client/python .
When you running docker run inside your script, kubernetes isn't aware of those containers, and they are just orphand containers running on your node host machine.