Skip to content

Policy Enforcement💣

Big Bang has several policies for Kubernetes resources to ensure best practices and security. For example, images must be pulled from Iron Bank, or containers must be run as non-root. These policies are currently enforced by OPA Gatekeeper, which gets deployed as the first package in Big Bang.

When integrating your package, you must adhere to the policies that are enforced or your resources will be denied by the Kubernetes admission controller. The following is how to identify and fix policy violations.

Prerequisites💣

  • a K8s cluster with Big Bang installed.
  • cluster admin access to the cluster with kubectl.

Integration💣

1. Deploying a Policy Enforcement Tool (OPA Gatekeeper)💣

The policy enforcement tool is deployed as the first package in the default Big Bang configuration. This is so that the enforcement tool can effectively protect the cluster from the start. Your package will be deployed on top of the Big Bang enforcement tool. The policy enforcement tool will control your package’s access to the cluster.

2. Identifying Violations Found on Your Application💣

In the following section, you will be shown how to identify violations found in your package. The app PodInfo will be used for all of the examples. Gatekeeper has three enforcement actions deny, dryrun, and warn. Only deny will prohibit access to the cluster, but the warn and dryrun constraints should be fixed as well as they are generally best practice.

In this example we will be attempting to install PodInfo onto our cluster:

 helm install flux-podinfo chart
NAME: flux-podinfo
LAST DEPLOYED: Mon Feb 14 11:24:26 2022
NAMESPACE: default
STATUS: deployed
REVISION: 1
NOTES:
1. Get the application URL by running these commands:
  echo "Visit http://127.0.0.1:8080 to use your application"
  kubectl -n default port-forward deploy/flux-podinfo 8080:9898

Everything looks good with the deployment, but upon further inspection we can see that our app hasn’t deployed properly.

 kubectl get all -n default
NAME                   TYPE        CLUSTER-IP   EXTERNAL-IP   PORT(S)             AGE
service/kubernetes     ClusterIP   10.43.0.1    <none>        443/TCP             52m
service/flux-podinfo   ClusterIP   10.43.39.6   <none>        9898/TCP,9999/TCP   20s

NAME                           READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/flux-podinfo   0/1     0            0           19s

NAME                                      DESIRED   CURRENT   READY   AGE
replicaset.apps/flux-podinfo-84d5bccfd6   1         0         0       19s

In order to get more information on why our deployment isn’t available, we can check the events of the K8s cluster. This will show us if there are policy violations, but will also reveal any other issues in our cluster.

  kubectl get events -n default
NAMESPACE  LAST SEEN  TYPE     REASON        OBJECT                              MESSAGE
default    31s        Warning  FailedCreate  replicaset/flux-podinfo-84d5bccfd6  Error creating: admission webhook "validation.gatekeeper.sh" denied the request: [no-privileged-containers] Privileged container is not allowed: podinfo, securityContext: {"privileged": true}

We can see that the issue in our example is that our PodInfo application is running containers as privileged.

To get more information as to how to fix this issue we can get the logs of the gatekeeper control plane

This is going to output a lot of logs to sift through so we can do a simple grep command looking for the specific policy denial, in this case no-privileged-containers.

kubectl logs -l control-plane=controller-manager -n gatekeeper-system --tail=-1 | grep "no-privileged-containers"

And we’ll see one of the log lines will looks something like the following:

{
  "level": "info",
  "ts": 1645018228.7589638,
  "logger": "webhook",
  "msg": "denied admission",
  "process": "admission",
  "event_type": "violation",
  "constraint_name": "no-privileged-containers",
  "constraint_group": "constraints.gatekeeper.sh",
  "constraint_api_version": "v1beta1",
  "constraint_kind": "K8sPSPPrivilegedContainer2",
  "constraint_action": "deny",
  "resource_group": "",
  "resource_api_version": "v1",
  "resource_kind": "Pod",
  "resource_namespace": "default",
  "resource_name": "flux-podinfo-84d5bccfd6-4l6tq",
  "request_username": "system:serviceaccount:kube-system:replicaset-controller"
}

3. Fixing Policy Violations💣

We can see the constraint_action: deny indicates that our resource was denied access to the cluster. The constraint_name and constraint_kind can provide us a way to get more information as to why our resource was denied. Running the following command will help you do so.

kubectl get <constraint_kind>.constraints.gatekeeper.sh/<constraint_name> -o json | jq '.metadata.annotations'

Replacing the command with our information give us the following:

kubectl get K8sPSPPrivilegedContainer2.constraints.gatekeeper.sh/no-privileged-containers -o json | jq '.metadata.annotations'
{
  "constraints.gatekeeper/description": "Containers must not run as privileged.",
  "constraints.gatekeeper/docs": "https://kubernetes.io/docs/concepts/workloads/pods/#privileged-mode-for-containers",
  "constraints.gatekeeper/name": "Privileged Containers",
  "constraints.gatekeeper/source": "https://github.com/open-policy-agent/gatekeeper-library/tree/master/library/pod-security-policy/privileged-containers",
  "helm.sh/hook": "post-install,post-upgrade"
}

The annotations provide us documentation information for specific policies as well as the source code to view that policy.

To fix this issue, navigate to your package’s chart/values.yaml or deployment.yaml and remove privileged: true or explicitly set it to false.

4. Exemptions to Policy Exceptions💣

Fixing the violation in the application is preferred, but sometimes we need to make an exception to the policy and leave the violation in place.

If you require an exception to a policy, please reference our exception doc for more information.

Validation💣

After we fixed the violation, we can run helm upgrade flux-podinfo chart. We can now check all the events in our cluster. This will show us if we’ve fixed our policy violation, but will also reveal non-policy related issues.

kubectl get all -n default

Last update: 2022-08-18 by Micah Nagel