Erratic and slow behavior with create-react-app and Skaffold kubernetes - reactjs

Question

I have Skaffold working well with local development server and database deployments. I'm trying to get working on the create-react-app front-end, but the behavior is incredibly slow and erratic.
Issues
The main problems are the following:
It takes upwards of five minutes from running skaffold dev --port-forward --tail for it finally start spinning up. Running just a docker build takes less than 30 seconds.
When it finally starts spinning up, it just sits on Starting the development server... for another two minutes.
Then, nine times out of ten, I get the following errors after several minutes (there are three because that is how many replicas there are):
One out of ten times, it will actually go into the Compiled Successfully! You can now view in the browser. It never does launch in Chrome though.
Changes to JS in create-react-app are never reflected in new browser. You have to stop and run Skaffold again. Skaffold does say Syncing 1 files for <image>... Watching for changes..., but nothing changes even after a refresh.
What I've tried
I've really simplified what I'm trying to do to make it easier to sort this out, so I'm using just an OOTB create-react-app application. The behavior is the same regardless.
minikube delete and minikube start several times (did this because even the server deployment started acting erratically after trying create-react-app)
Code and Steps to Reproduce
I'm on macOS Mojave (10.14.6) using Docker for Mac, Kubernetes (v1.16.0), minikube (v1.4.0), Skaffold (v0.39.0), and create-react-app. I'll have to skip the installation process for all of these since it is fairly lengthy, so the following steps assume you have this already setup.
Make a project directory:
mkdir project
Make a Kubernetes manifest directory and move into it:
mkdir k8s && cd k8s
Make a client-deployment.yaml and add the following:
apiVersion: apps/v1
kind: Deployment
metadata:
name: client-deployment
spec:
replicas: 3
selector:
matchLabels:
component: web
template:
metadata:
labels:
component: web
spec:
containers:
- name: client
image: testapp/client
ports:
- containerPort: 3000
Make a client-cluster-ip-service.yaml and add the following:
apiVersion: v1
kind: Service
metadata:
name: client-cluster-ip-service
spec:
type: ClusterIP
selector:
component: web
ports:
- port: 3000
targetPort: 3000
Move back into the parent:
cd ..
Create a skaffold.yaml and add the following:
apiVersion: skaffold/v1beta15
kind: Config
build:
local:
push: false
artifacts:
- image: testapp/client
context: web
docker:
dockerfile: Dockerfile.dev
sync:
manual:
- src: "**/*.js"
dest: .
- src: "**/*.html"
dest: .
- src: "**/*.css"
dest: .
deploy:
kubectl:
manifests:
- k8s/client-deployment.yaml
- k8s/client-cluster-ip-service.yaml
portForward:
- resourceType: service
resourceName: client-cluster-ip-service
port: 3000
localPort: 3000
Start a new create-react-app project:
npx create-react-app test-app
Change into the directory:
cd test-app
Create a Dockerfile.dev and add the following:
FROM node:alpine
WORKDIR '/app'
EXPOSE 3000
CMD ["npm", "run", "start"]
COPY package* ./
RUN npm install
COPY . .
Create a .dockerignore file and add the following:
node_modules
*.swp
Go back into the parent directory:
cd ..
Make sure minikube is running:
minikube start
Run the skaffold.yaml:
skaffold dev --port-forward --tail
This is what produces the issues for me.

Answer 1

Ok. Disregard. Started with one replica and it worked fine. Two worked fine. Three worked if skaffold was already running, but not from a fresh skaffold dev --port-forward --tail.
skaffold ssh and then did a top. Was running out of RAM... well was at 86% utilization. Increased it from the default 2GB to 8GB and now it works fine.
First deleted the VM with minikube delete and then created a new one with minikube start --memory='8g'. All good now.