rinaldodev / cos-fleetshard-1

The cos-fleetshard-operator responsible for provisioning and managing connectors instances on a cluster.

cos-fleetshard

set-up

• requirements:

  • jq

  • curl

  • stern

  • ocm

  • kubectl

  • uuidgen

  • cos-tools

  • minikube

• define the following env vars:

  • COS_BASE_PATH → base URL for the managed connector service control plane

  • KAS_BASE_PATH → base URL for the managed kafka service control plane

    Tip	I use direnv with the following set-up export OCM_CONFIG=$PWD/.ocm.json export KUBECONFIG=$PWD/.kube/config export COS_BASE_PATH=https://cos-fleet-manager-cos.rh-fuse-153f1de160110098c1928a6c05e19444-0000.eu-de.containers.appdomain.cloud export KAS_BASE_PATH=https://api.openshift.com

• retrieve your ocm-offline-token from https://qaprodauth.cloud.redhat.com/openshift/token using the _kafka_supporting account

  • follow the steps on that page to download and install the ocm command-line tool, then run the ocm login with the provided token

local installation

Note	This is an example installation that consists in: 1 sync 1 camel connector operator 1 debezium connector operator

• set-up minikube

  ./etc/scripts/start_minikube.sh

• install images

  eval $(minikube --profile cos docker-env)
  ./mvnw clean install -DskipTests=true -Pcontainer-build

• install the service

  kubectl apply -k etc/kubernetes/manifests/overlays/local

  If the previous command ends with an error like error: unable to recognize "etc/kubernetes/manifests/overlays/local": no matches for kind "IntegrationPlatform" in version "camel.apache.org/v1", then run the following:

  kubectl apply -f etc/kubernetes/manifests/overlays/local/camel-k/integration-platform.yaml

  At this point, operators and sync are deployed in the redhat-openshift-connectors, but they are not running as replica is set to 0 by default because some resources have to be configured.

  Point to the cos namespace

  kubectl config set-context --current --namespace=redhat-openshift-connectors

  ➜ kubectl get deployments -n redhat-openshift-connectors
  NAME                               READY   UP-TO-DATE   AVAILABLE   AGE
  camel-k-operator                   1/1     1            1           2d3h
  cos-fleetshard-operator-camel      0/0     0            0           6s
  cos-fleetshard-operator-debezium   0/0     0            0           5s
  cos-fleetshard-sync                0/0     0            0           4s
  strimzi-cluster-operator           1/1     1            1           2d3h

• configure pull secret

  In order to use private image on quay a pull secret need to be crated:

  • copy the content of rhoas-pull-docker to a local file

  • create a pull secret:

    kubectl create secret generic addon-pullsecret \
        --from-file=.dockerconfigjson=${path of rhoas-pull-docker} \
        --type=kubernetes.io/dockerconfigjson

• create cluster and configure secrets

  This section expects you to have cos-tools/bin in your PATH, but you may also just run the scripts from inside the bin directory.

  Note	This creates a new cluster for you on the fleet manager, remember to delete it once done.

  SUFFIX=$(uuidgen | tr -d '-')
  create-cluster-secret $(create-cluster "$USER-$SUFFIX" | jq -r '.id')

  When you’re done you may query for created clusters with get-clusters and delete it with delete-clusters <cluster id>.

• scale deployments

  kubectl scale deployment -l "app.kubernetes.io/part-of=cos" --replicas=1

• uninstall the service

  kubectl delete -k etc/kubernetes/manifests/overlays/local

local development

You may want to use quarkus:dev for local development, or if you want to run connected to a local fleet-manager. In this case, do the following:

• stop running containers

  Scale to 0 the components you want to run locally. For e.g., if you want to run the sync:

  kubectl scale deployment cos-fleetshard-sync --replicas=0

• check env vars

  If you are also running the fleet-manager locally, you need to change COS_BASE_PATH accordingly:

  export COS_BASE_PATH=<your_fleet_manager>

• make sure secret is correct

  If you change the COS_BASE_PATH variable, you need to recreate the cluster secret:

  create-cluster-secret <your-cluster-id>

• run using local profile

  Use the local maven profile to run the application. It will use quarkus:dev under the hoods, use an available port, set our namespace, and also import properties values from kubernetes secrets and configmaps:

  mvn -Dlocal

testing

Note	Although this section expects you to use a completely new kubernetes cluster, you may also just stop camel-k-operator and strimzi-cluster-operator and run the tests if you are an experienced user.

• set-up minikube

  # you may need to tune this command
  ./etc/scripts/start_minikube.sh

• install CRDs

  # install custom resources
  ./etc/scripts/deploy_fleetshard_crds.sh
  ./etc/scripts/deploy_camel-k_crds.sh
  ./etc/scripts/deploy_strimzi_crds.sh

• run tests

  ./mvnw clean install

components

• sync

• camel

• debezium