Migrating to version 1.0 of the Datadog Operator

The v1alpha1 DatadogAgent reconciliation in the Operator is deprecated since v1.2.0+ and will be removed in v1.7.0. After it's removed, you will not be able to configure the Datadog Operator to reconcile the v1alpha1 DatadogAgent CRD. However, you will still be able to apply a v1alpha1 manifest with the conversion webhook enabled using datadogCRDs.migration.datadogAgents.conversionWebhook.enabled.
DatadogAgent v1alpha1 and the conversion webhook will be removed in v1.8.0. After it's removed, you will not be able to migrate unless you use earlier version of the Operator.

Datadog Operator v0.X uses v1alpha1 of the DatadogAgent custom resource. Datadog Operator v1.X reconciles v2alpha1.

This guide describes how to migrate to the v2alpha1/DatadogAgent custom resource from v1alpha1/DatadogAgent.

Requirements

If you are using v1alpha1 with a 0.X version of the Datadog Operator and would like to upgrade, you need to use the conversion webhook feature.

Start by ensuring that you have the minimum required version of the chart and its dependencies:

NAME                	CHART VERSION	APP VERSION	DESCRIPTION
datadog/datadog-crds	1.0.0        	1          	Datadog Kubernetes CRDs chart

For the Datadog Operator chart:

NAME                    	CHART VERSION	APP VERSION	DESCRIPTION
datadog/datadog-operator	1.0.0        	1.0.0      	Datadog Operator

Install cert manager

If you do not already have the cert manager, install it with Helm.

Add the chart:

helm repo add jetstack https://charts.jetstack.io

Then, install it:

 helm install \
  cert-manager jetstack/cert-manager \
  --version v1.11.0 \
  --set installCRDs=true

Migration

Run the following command to redeploy the Datadog Operator and configure Kubernetes to store version v2alpha1 of the DatadogAgent:

helm upgrade \
    datadog-operator datadog/datadog-operator \
    --set image.tag=1.0.0 \
    --set datadogCRDs.migration.datadogAgents.version=v2alpha1 \
    --set datadogCRDs.migration.datadogAgents.useCertManager=true \
    --set datadogCRDs.migration.datadogAgents.conversionWebhook.enabled=true

With this, the conversion webhook server (run by the Datadog Operator) converts existing DatadogAgent objects.

If you have v1alpha1 versions and migrate, it is recommended that you save the converted version and start solely deploying the converted version. Once you deploy only the v2alpha1 DatadogAgent, you can disable the conversion webhook.

Notes

Starting at the version 1.0.0 of the datadog-operator chart, the field image.tag has a default values of 1.0.0, and datadogCRDs.migration.datadogAgents.version is v2alpha1.

These are set in the command here to illustrate the migration of going from a Datadog Operator version < 1.0.0 with a stored version of v1alpha1 to the GA version of 1.0.0 with a stored version of v2alpha1.

Implementation details

This creates a self-signed Certificate (using an Issuer) that is used by the Certificate Manager to mutate the DatadogAgent CRD to document the caBundle that the API Server uses to contact the conversion webhook.

The Datadog Operator is running the reconciler for v2alpha1 object and starts a conversion webhook server, exposed on port 9443. The API Server uses this server to convert v1alpha1 DatadogAgent to v2alpha1.

Lifecycle

The conversion webhook is not meant to run indefinitely. Datadog only recommends it to migrate your objects during a transitional period.

Once converted, you can store the new version of your DatadogAgent, deactivate the conversion, and deploy only v2alpha1 objects.

Troubleshooting

I don’t see the v2alpha1 version of the DatadogAgent resource.

Because v1alpha1 and v2alpha1 are served, you might need to specify which version you want to see:

kubectl get datadogagents.v2alpha1.datadoghq.com datadog-agent

The conversion is not working.

The logs of the Datadog Operator pod should show that the conversion webhook is enabled, the server is running, and the certificates are watched.

kubectl logs datadog-operator-XXX-YYY
[...]
{"level":"INFO","ts":"2023-02-16T16:47:07Z","logger":"controller-runtime.webhook","msg":"Registering webhook","path":"/convert"}
{"level":"INFO","ts":"2023-02-16T16:47:07Z","logger":"controller-runtime.builder","msg":"Conversion webhook enabled","GVK":"datadoghq.com/v2alpha1, Kind=DatadogAgent"}
{"level":"INFO","ts":"2023-02-16T16:47:07Z","logger":"setup","msg":"starting manager"}
{"level":"INFO","ts":"2023-02-16T16:47:07Z","logger":"controller-runtime.webhook.webhooks","msg":"Starting webhook server"}
{"level":"INFO","ts":"2023-02-16T16:47:07Z","logger":"controller-runtime.certwatcher","msg":"Updated current TLS certificate"}
{"level":"INFO","ts":"2023-02-16T16:47:07Z","logger":"controller-runtime.webhook","msg":"Serving webhook server","host":"","port":9443}
{"level":"INFO","ts":"2023-02-16T16:47:07Z","msg":"Starting server","path":"/metrics","kind":"metrics","addr":"0.0.0.0:8383"}
{"level":"INFO","ts":"2023-02-16T16:47:07Z","msg":"Starting server","kind":"health probe","addr":"0.0.0.0:8081"}
{"level":"INFO","ts":"2023-02-16T16:47:07Z","logger":"controller-runtime.certwatcher","msg":"Starting certificate watcher"}
[...]

How do I check the service registered for the conversion for a registered endpoint?

kubectl describe service datadog-operator-webhook-service
[...]
Name:              datadog-operator-webhook-service
Namespace:         default
[...]
Selector:          app.kubernetes.io/instance=datadog-operator,app.kubernetes.io/name=datadog-operator
[...]
Port:              <unset>  443/TCP
TargetPort:        9443/TCP
Endpoints:         10.88.3.28:9443

How do I verify the registered service for the conversion webhook?

kubectl describe crd datadogagents.datadoghq.com
[...]
  Conversion:
    Strategy:  Webhook
    Webhook:
      Client Config:
        Ca Bundle:  LS0t[...]UtLS0tLQo=
        Service:
          Name:       datadog-operator-webhook-service
          Namespace:  default
          Path:       /convert
          Port:       443
      Conversion Review Versions:
        v1

The CRD does not have the caBundle.

Make sure that the CRD has the correct annotation: cert-manager.io/inject-ca-from: default/datadog-operator-serving-cert. Also, check the logs of the cert-manager-cainjector pod.

If you do not see anything standing out, setting the log level to 5 (debug) might help:

kubectl edit deploy cert-manager-cainjector -n cert-manager
[...]
    spec:
      containers:
      - args:
        - --v=5
[...]

You should see logs such as:

[...]
I0217 08:11:15.582479       1 controller.go:178] cert-manager/certificate/customresourcedefinition/generic-inject-reconciler "msg"="updated object" "resource_kind"="CustomResourceDefinition" "resource_name"="datadogagents.datadoghq.com" "resource_namespace"="" "resource_version"="v1"
I0217 08:25:24.989209       1 sources.go:98] cert-manager/certificate/customresourcedefinition/generic-inject-reconciler "msg"="Extracting CA from Certificate resource" "certificate"="default/datadog-operator-serving-cert" "resource_kind"="CustomResourceDefinition" "resource_name"="datadogagents.datadoghq.com" "resource_namespace"="" "resource_version"="v1"
[...]

Rollback

If you migrated to the new version of the Datadog Operator using v2alpha1 but want to roll back to the former version, Datadog recommends:

  • Scaling the Datadog Operator deployment to 0 replicas.
    kubectl scale deploy datadog-operator --replicas=0
    
  • Upgrading the chart to have v1alpha1 stored and for the Datadog Operator to use the 0.8.X image.
    helm upgrade \
      datadog-operator datadog/datadog-operator \
      --set image.tag=0.8.4 \
      --set datadogCRDs.migration.datadogAgents.version=v1alpha1 \
      --set datadogCRDs.migration.datadogAgents.useCertManager=false \
      --set datadogCRDs.migration.datadogAgents.conversionWebhook.enabled=false
    
  • Redeploy the previous DatadogAgent v1alpha1 object.

Note: The DaemonSet of the Datadog Agents is rolled back in the process.

PREVIEWING: rtrieu/product-analytics-ui-changes