Files
falco/examples/k8s_audit_config

This page describes how to get Kubernetes Auditing working with Falco. Either using static audit backends in Kubernetes 1.11, or in Kubernetes 1.13 with dynamic sink which configures webhook backends through an AuditSink API object.

Instructions for Kubernetes 1.11

The main steps are:

  1. Deploy Falco to your Kubernetes cluster
  2. Define your audit policy and webhook configuration
  3. Restart the API Server to enable Audit Logging
  4. Observe Kubernetes audit events at falco

Deploy Falco to your Kubernetes cluster

Follow the Kubernetes Using Daemonset instructions to create a falco service account, service, configmap, and daemonset.

Define your audit policy and webhook configuration

The files in this directory can be used to configure Kubernetes audit logging. The relevant files are:

  • audit-policy.yaml: The Kubernetes audit log configuration we used to create the rules in k8s_audit_rules.yaml.
  • webhook-config.yaml.in: A (templated) webhook configuration that sends audit events to an ip associated with the falco service, port 8765. It is templated in that the actual IP is defined in an environment variable FALCO_SERVICE_CLUSTERIP, which can be plugged in using a program like envsubst.

Run the following to fill in the template file with the ClusterIP IP address you created with the falco-service service above. Although services like falco-service.default.svc.cluster.local can not be resolved from the kube-apiserver container within the minikube vm (they're run as pods but not really a part of the cluster), the ClusterIPs associated with those services are routable.

FALCO_SERVICE_CLUSTERIP=$(kubectl get service falco-service -o=jsonpath={.spec.clusterIP}) envsubst < webhook-config.yaml.in > webhook-config.yaml

Restart the API Server to enable Audit Logging

A script enable-k8s-audit.sh performs the necessary steps of enabling audit log support for the apiserver, including copying the audit policy/webhook files to the apiserver machine, modifying the apiserver command line to add --audit-log-path, --audit-policy-file, etc. arguments, etc. (For minikube, ideally you'd be able to pass all these options directly on the minikube start command line, but manual patching is necessary. See this issue for more details.)

It is run as bash ./enable-k8s-audit.sh <variant> static. <variant> can be one of the following:

  • minikube
  • kops

When running with variant equal to kops, you must either modify the script to specify the kops apiserver hostname or set it via the environment: APISERVER_HOST=api.my-kops-cluster.com bash ./enable-k8s-audit.sh kops

Its output looks like this:

$ bash enable-k8s-audit.sh minikube static
***Copying apiserver config patch script to apiserver...
apiserver-config.patch.sh                                                                   100% 1190     1.2MB/s   00:00
***Copying audit policy/webhook files to apiserver...
audit-policy.yaml                                                                           100% 2519     1.2MB/s   00:00
webhook-config.yaml                                                                         100%  248   362.0KB/s   00:00
***Modifying k8s apiserver config (will result in apiserver restarting)...
***Done!
$

Observe Kubernetes audit events at falco

Kubernetes audit events will then be routed to the falco daemonset within the cluster, which you can observe via kubectl logs -f $(kubectl get pods -l app=falco-example -o jsonpath={.items[0].metadata.name}).

Instructions for Kubernetes 1.13

The main steps are:

  1. Deploy Falco to your Kubernetes cluster
  2. Restart the API Server to enable Audit Logging
  3. Deploy the AuditSink object for your audit policy and webhook configuration
  4. Observe Kubernetes audit events at falco

Deploy Falco to your Kubernetes cluster

Follow the Kubernetes Using Daemonset instructions to create a Falco service account, service, configmap, and daemonset.

Restart the API Server to enable Audit Logging

A script enable-k8s-audit.sh performs the necessary steps of enabling dynamic audit support for the apiserver by modifying the apiserver command line to add --audit-dynamic-configuration, --feature-gates=DynamicAuditing=true, etc. arguments, etc. (For minikube, ideally you'd be able to pass all these options directly on the minikube start command line, but manual patching is necessary. See this issue for more details.)

It is run as bash ./enable-k8s-audit.sh <variant> dynamic. <variant> can be one of the following:

  • minikube
  • kops

When running with variant equal to kops, you must either modify the script to specify the kops apiserver hostname or set it via the environment: APISERVER_HOST=api.my-kops-cluster.com bash ./enable-k8s-audit.sh kops

Its output looks like this:

$ bash enable-k8s-audit.sh minikube dynamic
***Copying apiserver config patch script to apiserver...
apiserver-config.patch.sh                                                                   100% 1190     1.2MB/s   00:00
***Modifying k8s apiserver config (will result in apiserver restarting)...
***Done!
$

Deploy AuditSink objects

audit-sink.yaml.in, in this directory, is a template audit sink configuration that defines the dynamic audit policy and webhook to route Kubernetes audit events to Falco.

Run the following to fill in the template file with the ClusterIP IP address you created with the falco-service service above. Although services like falco-service.default.svc.cluster.local can not be resolved from the kube-apiserver container within the minikube vm (they're run as pods but not really a part of the cluster), the ClusterIPs associated with those services are routable.

FALCO_SERVICE_CLUSTERIP=$(kubectl get service falco-service -o=jsonpath={.spec.clusterIP}) envsubst < audit-sink.yaml.in > audit-sink.yaml

Observe Kubernetes audit events at falco

Kubernetes audit events will then be routed to the falco daemonset within the cluster, which you can observe via kubectl logs -f $(kubectl get pods -l app=falco-example -o jsonpath={.items[0].metadata.name}).

Instructions for Kubernetes 1.13 with dynamic webhook and local log file

If you want to use a mix of AuditSink for remote audit events as well as a local audit log file, you can run enable-k8s-audit.sh with the "dynamic+log" argument e.g. bash ./enable-k8s-audit.sh <variant> dynamic+log. This will enable dynamic audit logs as well as a static audit log to a local file. Its output looks like this:

***Copying apiserver config patch script to apiserver...
apiserver-config.patch.sh                                                                          100% 2211   662.9KB/s   00:00
***Copying audit policy file to apiserver...
audit-policy.yaml                                                                                  100% 2519   847.7KB/s   00:00
***Modifying k8s apiserver config (will result in apiserver restarting)...
***Done!

The audit log will be available on the apiserver host at /var/lib/k8s_audit/audit.log.