This is just experiments with:

• Go (v1.15.1)
• Kubernetes (OSX Docker Desktop)
• Helm (v2.9.0+, v3.+)
  • environment yaml files to env variables
  • environment global with local overrides

# Build the docker image
./bin/docker-build-push.sh -n NAME -v VERSION
# Update Chart.yaml appVersion to the VERSION above
# helm deploy 
./bin/deploy.sh

Environment variables are exposed to the service from YAML files in environment folder. The YAML, containing a map named config is mapped to environment variables as follows:

config:
    max: "42"
    database:
      name: "server.name.com"
      port: "2056"

The environment variables that result will be:

CONFIG_MAX=42
CONFIG_DATABASE_NAME=server.name.com
CONFIG_DATABASE_PORT=2056

# Remove 2.9.0
helm delete NAME --purge

# Remove 3.+
helm uninstall NAME

# Install 2.9.0
helm install CHART_PATH -name NAME

# INSTALL 3.+
helm install NAME CHART_PATH

• Helm 2.9.0 issues:
  • Helm init doesn't work, see helm-init.sh for fix.
  • Helm delete, even with --purge leaves running pod. Need t use kubectl to kill.
• Helm 3+:
  • helm uninstall seems to work properly
• I'm not using a two stage build with a Golang docker image.
  • Building a Linux amd64 binary and copying that into an Alpine image in anticipation of having to build from a very large source base in the future and wanting to avoid multi-stage build pattern.
• Helm chart stripped back as much as possible.