Add an OPERATOR_LOCAL_MODE environment variable to launch locally.

[jkneubuh]

This PR allows the operator to be launched locally without running within a Docker container / kubernetes cluster.

Fixes Issue #67

Note that this changes the default path at which the operator config files are read, from ../../ to ./. This was actually a bug / error in the original logic, as the working directory for ibp-operator in the container is actually /. (When running from the root folder, the ../../ path prefix resolves to / on the file system.)

Thank you @bjwswang for the great recommendations on pushing this forward.

Signed-off-by: Josh Kneubuhl jkneubuh@us.ibm.com

[jkneubuh]

@bjwswang does this PR enable you to run the operator locally without deploying to a container in k8s?

The env necessary to run the operator is:

export KUBECONFIG=$HOME/.kube/config
 export WATCH_NAMESPACE=test-network
 export CLUSTERTYPE=k8s
 export OPERATOR_LOCAL_MODE=true

The PWD for the process must be the root of the git project.

Thank you very much for opening the issue and submitting a fix for the defect. This feature has been waiting for a VERY long time...

[bjwswang]

Hi @jkneubuh looking good for running locally,but how about docker mode? From the Dockerfile

COPY build/ /usr/local/bin
COPY definitions /definitions
COPY config/crd/bases /deploy/crds
COPY defaultconfig /defaultconfig
COPY docker-entrypoint.sh .

the operator executable file placed under /usr/local/bin,so not sure if relative path ./definitionsstill works on docker mode. maybe we make the base path of default configuration/template files to be configurable.

[jkneubuh]

When running in a docker container, the $PWD for the process is /. I think the older default path prefix of ../../ was incorrect - it was resolving correctly by accident as any ancestor path of '/' will resolve to '/'. E.g.:

fuji:fabric-operator jkneubuh$ docker run -e LICENSE=accept --rm ghcr.io/hyperledger-labs/fabric-operator:latest-amd64 pwd 
/
fuji:fabric-operator jkneubuh$ docker run -e LICENSE=accept --rm ghcr.io/hyperledger-labs/fabric-operator:latest-amd64 ls ./defaultconfig
ca
console
orderer
peer
fuji:fabric-operator jkneubuh$ docker run -e LICENSE=accept --rm ghcr.io/hyperledger-labs/fabric-operator:latest-amd64 ls ../../defaultconfig
ca
console
orderer
peer
fuji:fabric-operator jkneubuh$ docker run -e LICENSE=accept --rm ghcr.io/hyperledger-labs/fabric-operator:latest-amd64 ls ../../../../../../../defaultconfig
ca
console
orderer
peer

I tested this with a container / k8s and it looks OK. To test this in k8s without turning a real build through CI, I did:

make image 
docker tag docker tag  ghcr.io/hyperledger-labs/fabric-operator:latest-amd64 localhost:5000/fabric-operator 
docker push localhost:5000/fabric-operator

export TEST_NETWORK_OPERATOR_IMAGE=localhost:5000/fabric-operator
export TEST_NETWORK_OPERATOR_IMAGE_LABEL=latest

network operator