Article

Set up Red Hat AMQ Streams custom certificates on OpenShift

December 18, 2019
Kubernetes
Federico Valeri
Senior Software Engineer

Secure communication over a computer network is one of the most important requirements for a system, and yet it can be difficult to set up correctly. This example shows how to set up Red Hat AMQ Streams' end-to-end TLS encryption using a custom X.509 CA certificate on the Red Hat OpenShift platform.

Prerequisites

You need to have the following in place before you can proceed with this example:

The procedure

Before we start, let's define a few handy variables: 

USER="developer"
PROJECT="streams"
CA_USER="system:admin"
RA_SECRET="reg-auth-secret"
CLUSTER="my-cluster"

Set up a new project

The first step after this is to log in as cluster-admin and create a new project to host our clusters. We need this role because we have to install custom resource definitions (CRDs) that are required by the Cluster Operator (CO). We then give full admin rights to the user to let them manage the project once ready: 

$ oc login -u $CA_USER
$ oc new-project $PROJECT
$ oc adm policy add-role-to-user admin $USER

To be able to download images from the Red Hat Container Registry, we also need to add an authentication Secret (use your credentials here): 

$ oc create secret docker-registry $RA_SECRET \
      --docker-server=registry.redhat.io \
      --docker-username=<portal-username> \
      --docker-password=<portal-password>
$ oc secrets link default $RA_SECRET --for=pull

Then, unzip the Installation and Examples distribution package and replace the default project's name with yours: 

TMP="/tmp/$PROJECT" && rm -rf $TMP && mkdir -p $TMP
$ unzip -qq amq-streams-1.3.0-ocp-install-examples.zip -d $TMP
$ sed -i -e "s/namespace: .*/namespace: $PROJECT/g" $TMP/install/cluster-operator/*RoleBinding*.yaml

Now, we are ready to install all required CRDs and the Strimzi CO: 

$ oc apply -f $TMP/install/cluster-operator
$ oc secrets link strimzi-cluster-operator $RA_SECRET --for=pull
$ oc set env deploy/strimzi-cluster-operator STRIMZI_IMAGE_PULL_SECRETS=$RA_SECRET

$ oc set env deploy/strimzi-cluster-operator STRIMZI_NAMESPACE=$PROJECT
$ oc apply -f $TMP/install/cluster-operator/020-RoleBinding-strimzi-cluster-operator.yaml
$ oc apply -f $TMP/install/cluster-operator/032-RoleBinding-strimzi-cluster-operator-topic-operator-delegation.yaml
$ oc apply -f $TMP/install/cluster-operator/031-RoleBinding-strimzi-cluster-operator-entity-operator-delegation.yaml
$ oc apply -f $TMP/install/strimzi-admin
$ oc adm policy add-cluster-role-to-user strimzi-admin $USER

Configure the custom certificate

After these commands finish, we can configure our custom X.509 CA certificate. I expect that you already have the following files:

  • rootca.pem: The root Certificate Authority (CA) in our domain (optional).
  • intermca.pem: An intermediate CA used to sign the certificate in a specific context (optional).
  • myca.pem: Our custom CA certificate to use with Apache Kafka.
  • myca-prk.pem: The private key for our custom CA certificate.

All CAs in the chain should be configured as a CA in the X509v3 Basic Constraints. This means that you cannot use a classic non-CA certificate to replace the self-generated certificate (see also additional notes at the end). The reason for this is that it is used to sign certificates for inter-broker communication.

After printing out your custom certificate you should be able to see this property: 

$ openssl x509 -inform pem -in myca.pem -noout -text
...
X509v3 Basic Constraints: 
    CA:TRUE

When you have a valid CA certificate, create a bundle file like this: 

$ cat myca.pem intermca.pem rootca.pem > bundle.pem

Then, create all required Secrets and labels containing our custom CA. This must be done before creating our custom cluster (next step): 

$ oc create secret generic $CLUSTER-cluster-ca-cert --from-file=ca.crt=bundle.pem
$ oc label secret $CLUSTER-cluster-ca-cert strimzi.io/kind=Kafka strimzi.io/cluster=$CLUSTER

$ oc create secret generic $CLUSTER-cluster-ca --from-file=ca.key=myca-prk.pem
$ oc label secret $CLUSTER-cluster-ca strimzi.io/kind=Kafka strimzi.io/cluster=$CLUSTER

$ oc create secret generic $CLUSTER-clients-ca-cert --from-file=ca.crt=bundle.pem
$ oc label secret $CLUSTER-clients-ca-cert strimzi.io/kind=Kafka strimzi.io/cluster=$CLUSTER

$ oc create secret generic $CLUSTER-clients-ca --from-file=ca.key=myca-prk.pem
$ oc label secret $CLUSTER-clients-ca strimzi.io/kind=Kafka strimzi.io/cluster=$CLUSTER

Finally, we can deploy our cluster definition. Note how we set generateCertificateAuthority to instruct the CO not to generate the self-signed CA that otherwise would overwrite our previous configuration.

Example: Ephemeral cluster creation (not for production)

Here we create a small ephemeral cluster just for the sake of this example. Do not use the exact same setup for production: 

$ oc create -f - <<EOF
apiVersion: kafka.strimzi.io/v1alpha1
kind: Kafka
metadata:
  name: $CLUSTER
spec:
  kafka:
    version: "2.3.0"
    replicas: 1
    config:
      num.partitions: 1
      default.replication.factor: 1
      log.message.format.version: "2.3"
    clusterCa:
      generateCertificateAuthority: false
    clientsCa:
      generateCertificateAuthority: false
    listeners:
      plain: {}
      tls: {}
      external:
        type: route
    readinessProbe:
      initialDelaySeconds: 30
      timeoutSeconds: 10
    livenessProbe:
      initialDelaySeconds: 30
      timeoutSeconds: 10
    template:
        pod:
          terminationGracePeriodSeconds: 120
    storage:
      type: ephemeral
    resources:
      requests:
        cpu: "1000m"
        memory: "2Gi"
      limits:
        cpu: "1000m"
        memory: "2Gi"
    tlsSidecar:
      resources:
        limits:
          cpu: "100m"
          memory: "128Mi"
        requests:
          cpu: "100m"
          memory: "128Mi"
  zookeeper:
    replicas: 1
    readinessProbe:
      initialDelaySeconds: 15
      timeoutSeconds: 5
    livenessProbe:
      initialDelaySeconds: 15
      timeoutSeconds: 5
    storage:
      type: ephemeral
    resources:
      requests:
        cpu: "500m"
        memory: "1Gi"
      limits:
        cpu: "500m"
        memory: "1Gi"
    tlsSidecar:
      resources:
        limits:
          cpu: "100m"
          memory: "128Mi"
        requests:
          cpu: "100m"
          memory: "128Mi"
  entityOperator:
    topicOperator:
      resources:
        limits:
          cpu: "250m"
          memory: "256Mi"
        requests:
          cpu: "250m"
          memory: "256Mi"
    userOperator:
      resources:
        limits:
          cpu: "250m"
          memory: "256Mi"
        requests:
          cpu: "250m"
          memory: "256Mi"
    tlsSidecar:
      resources:
        limits:
          cpu: "100m"
          memory: "128Mi"
        requests:
          cpu: "100m"
          memory: "128Mi"
EOF

Once the cluster is up and running, you might want to check that the custom CA is correctly loaded: 

$ oc get pods
$ oc logs strimzi-cluster-operator-<uuid>
$ oc logs $CLUSTER-kafka-0 -c kafka

Set up the Java client

Create and use a truststore in Java KeyStore (JKS) format for one-way TLS authentication: 

$ oc extract secret/$CLUSTER-cluster-ca-cert --keys=ca.crt --to=- > ca.pem
keytool -import -noprompt -alias root -file ca.pem -keystore truststore.jks -storepass secret

If you want to access Kafka from outside OpenShift, then you also need to use this bootstrap URL: 

$ echo $(oc get routes $CLUSTER-kafka-bootstrap -o=jsonpath='{.status.ingress[0].host}{"\n"}'):443

Additional notes

We already know that most security teams won't easily release CA certificates. We are working on an enhancement to provide the option to use a non-CA certificate for Kafka listeners, leaving the internal self-generated CA to secure the inter-broker communication.

Beware that when using a custom CA as explained in this post, you are responsible for the certificate renewals. This process is fully automated when using self-generated certificates. In any case, after the renewal, you will have to recreate the client's truststore as described before.

Last updated: July 1, 2020

Comments