Documentation for version: v0.1.0

Developer guide

Operator

CI Jobs

  • Kind: E2E test running on each PR, can be found on .github/workflows/kind.yaml
  • Go/build: Builds current code and checks that unit-test are working
  • Go/Lint: Check that code is correctly linted
  • Go/YamlLint: Checks that Yaml files are correctly linted
  • Go/SecurityScanner: Checks that the current code is statically checked by gosec
  • Go/Test Coverage: Reports the unit test code coverage, fails if went down.

Logging

Log configuration can be found here

logr is a logging API for Go. It provides a simple interface for logging under which there is an actual logger that implements the logr interface.

Zap is a log library that implements logr interface.

SDK-generated operators use the logr interface to log. Operator SDK by default uses a zap-based logger that is ready for production use. The default verbosity is set to info level.

logr defines logger’s verbosity levels numerically. To write log lines that are more verbose, logr.Logger has a V() method. The higher the V-level of a log line, the less critical it is considered. Level V(0) is the default, and logger.V(0).Info() has the same meaning as logger.Info().

Levels in logr correspond to custom debug levels in Zap. Any given level in logr is represented by its inverse in zap (zapLevel = -1*logrLevel). Thus, in Zap, higher levels are more important.

For example: logr V(2) is equivalent to log level -2 in Zap, while logr V(1) is equivalent to debug level -1 in Zap.

To summarize:

Zap logging priority Zap enum logr
-1 debug .V(1).Info(...)
0 info .V(0).Info(...) or .Info(...)
1 warn N.A.
2 error .Error(...)
3 dpanic N.A.
4 panic N.A.
5 fatal N.A.

E2E tests

Prerequisites

  • k8s cluster or Openshift Cluster
  • Make sure that your oc or kubectl is configured properly to communicate with your cluster
  • Not needed for all test, OpenShift Cluster Storage or at least NooBaa.

Running the tests

  1. Create container image of flotta operator.
# Build the flotta operator
$ make build

# Make sure you have configured k8s provider docker
# For example for minikube by running: eval $(minikube -p minikube docker-env)
# Then run the build of the container
$ IMG=flotta-operator:latest make docker-build

# Deploy the operator on k8s
$ make deploy IMG=flotta-operator

# Wait until the operator is ready
$ kubectl wait --timeout=120s --for=condition=Ready pods -l app=flotta-controller-manager -n flotta
  1. Expose the flotta API
$ kubectl port-forward deploy/flotta-operator-controller-manager -n flotta --address 0.0.0.0 8043:8043 &
  1. Run the tests
$ make integration-test

Troubleshooting

kubectl wait timeouts:

If timeout, debug the deployment logs by running:

kubectl logs deploy/flotta-operator-controller-manager -n flotta

Waiting for edge device timeouts in integration tests:

Debug the edge device container by executing shell inside the container:

docker exec -it edgedevice1 journalctl -u yggdrasild.service

Device Worker

CI jobs

  • Go/build: Builds current code and checks that unittest are working
  • Go/Lint: Checks that code is correctly linted
  • Go/SecurityScanner: Checks that the current code is statically checked by gosec