Audit policy


This page describes the audit logging policy for Google Kubernetes Engine (GKE). For instructions to enable and view the different types of audit logs, as well as the required IAM permissions, refer to GKE audit logging information.

Before you begin

Before reading this page, ensure that you're familiar with these topics:

Overview

In a GKE cluster, the Kubernetes API server writes audit log entries to a backend that is managed by GKE. As GKE receives log entries from the Kubernetes API server, it writes them to your project's Admin Activity log and Data Access log.

There are two policies that influence what you see in your project's audit logs:

  • The Kubernetes audit policy defines rules for which events are recorded as log entries. It also specifies what data the log entries should include.

  • The GKE audit policy determines which entries are written to your Admin Activity log and which are written to your Data Access log.

The audit logs that are written to your Data Access log depend on your audit log configuration. Data Access logs use the Google Cloud Observability pricing. Admin Activity logs are offered at no charge. GKE supports the following types of Data Access logs.

  • ADMIN_READ: operations that read metadata about Kubernetes resources, such as listing Pods.
  • DATA_READ: operations that read data in Kubernetes resources, such as reading the logs for a Pod.
  • DATA_WRITE: operations that write data to Kubernetes resources, such as updating Pod status.

For more information, see Configure Data Access audit logs.

Kubernetes audit policy

The Kubernetes API server follows a policy that is specified in the --audit-policy-file flag of the kube-apiserver command.

When GKE starts the Kubernetes API server, it supplies an audit policy file by setting the value of the --audit-policy-file flag. In the open-source Kubernetes repository, you can see the configure-helper.sh script, which generates the audit policy file. You can see most of the audit policy file by looking directly at the script.

Events and stages

When a person or component makes a request to the Kubernetes API server, the request goes through one or more stages:

Stage Description
RequestReceived The audit handler has received the request.
ResponseStarted The response headers have been sent, but the response body has not been sent.
ResponseComplete The response body has been completed, and no more bytes will be sent.
Panic There was an internal server error, and the request did not complete..

Each stage of a request generates an event, which is processed according to a policy. The policy specifies whether the event should be recorded as a log entry and if so, what data should be included in the log entry.

Audit levels

The Kubernetes audit policy file contains a list of rules. In the policy file, the first rule that matches an event sets the audit level for the event.

A rule can specify one of these audit levels:

Level Description
None Do not create a log entry for the event.
Metadata Create a log entry. Include metadata, but do not include the request body or the response body.
Request Create a log entry. Include metadata and the request body, but do not include the response body.
RequestResponse Create a log entry. Include metadata, the request body, and the response body.

Here's an example of a rule. If an event matches the rule, the Kubernetes API server creates a log entry at the Request level.

- level: Request
  userGroups: ["system:nodes"]
  verbs: ["update","patch"]
  resources:
    - group: "" # core
      resources: ["nodes/status", "pods/status"]
  omitStages:
    - "RequestReceived"

An event matches the rule if all of the following are true:

  • The event does not match any previous rule in the policy file.
  • The identity making the call is in the system:nodes user group.
  • The call is an update request or a patch request.
  • The request is on a nodes/status resource or a pods/status resource.
  • The event is not for the RequestReceived stage of the call.

Summary of the Kubernetes audit policy for GKE clusters

  • In general, write requests like create, update, and delete are logged at the RequestResponse level.

  • In general, get, list, and watch events are logged at the Metadata level.

  • Some events are treated as special cases. For a definitive list of requests that are treated as special cases, see the policy script At the time of this writing, these are the special cases:

    • Update and patch requests by kubelet, system:node-problem-detector, or system:serviceaccount:kube-system:node-problem-detector on a nodes/status resource or a pods/status resource are logged at the Request level.

    • Update and patch requests by any identity in the system:nodes group on a nodes/status resource or a pods/status resource are logged at the Request level.

    • deletecollection requests by system:serviceaccount:kube-system:namespace-controller are logged at the Request level.

    • Requests on a secrets resource, a configmaps resource, or a tokenreviews resource are logged at the Metadata level.

  • Certain requests are not logged at all. For a definitive list of requests that are not logged, see the level: None rules in the policy script. At the time of this writing, these are the requests that don't get logged:

    • Requests by system:kube-proxy to watch an endpoints resource, a services resource, or a services/status resource.

    • Get requests by system:unsecured on a configmaps resource in the kube-system namespace.

    • Get requests by kubelet on a nodes resource or a nodes/status resource.

    • Get requests by any identity in the system:nodes group on a nodes resource or a nodes/status resource.

    • Get and update requests by system:kube-controller-manager, system:kube-scheduler, or system:serviceaccount:endpoint-controller on an endpoints resource in the kube-system namespace.

    • Get requests by system:apiserver on a namespaces resource, a namespaces/status resource, or a namespaces/finalize resource.

    • Get and list requests by system:kube-controller-manager on any resource in the metrics.k8s.io group.

    • Requests to URLs that match /healthz*, /version, or /swagger*.

GKE audit policy

As GKE receives log entries from the Kubernetes API server, it applies its own policy to determine which entries get written to your project's Admin Activity log and which entries get written to your project's Data Access log.

For the most part, GKE applies the following rules to log entries that come from the Kubernetes API server:

  • Entries that represent create, delete, and update requests go to your Admin Activity log.

  • Entries that represent get, list, and updateStatus requests go to your Data Access log.

For information about pricing and which types of logs are enabled by default, see Logging details.

Understanding the Kubernetes audit policy file

For GKE clusters, the Kubernetes audit policy file starts with rules that specify that certain events should not be logged at all. For example, this rule says that any get request by kubelet on a nodes resource or a nodes/status resource should not be logged. Recall that a level of None means that any matching event should not be logged:

- level: None
  users: ["kubelet"] # legacy kubelet identity
  verbs: ["get"]
  resources:
    - group: "" # core
    resources: ["nodes", "nodes/status"]

After the list of level: None rules, the policy file has a list of rules that are special cases. For example, here's a special-case rule says to log certain requests at the Metadata level:

- level: Metadata
    resources:
      - group: "" # core
        resources: ["secrets", "configmaps"]
      - group: authentication.k8s.io
        resources: ["tokenreviews"]
    omitStages:
      - "RequestReceived"

An event matches the rule if all of the following are true:

  • The event does not match any previous rule in the policy file.
  • The request is on a resource of type secrets, configmaps, or tokenreviews.
  • The event is not for the RequestReceived stage of the call.

After the list of special-case rules, the policy file has some general rules. To see the general rules in the script, you have to substitute the value of known_apis for ${known_apis}. After the substitution, you get a rule like this:

- level: Request
  verbs: ["get", "list", "watch"]
  resources:
    - group: "" # core
    - group: "admissionregistration.k8s.io"
    - group: "apiextensions.k8s.io"
    - group: "apiregistration.k8s.io"
    - group: "apps"
    - group: "authentication.k8s.io"
    - group: "authorization.k8s.io"
    - group: "autoscaling"
    - group: "batch"
    - group: "certificates.k8s.io"
    - group: "extensions"
    - group: "metrics.k8s.io"
    - group: "networking.k8s.io"
    - group: "policy"
    - group: "rbac.authorization.k8s.io"
    - group: "settings.k8s.io"
    - group: "storage.k8s.io"
  omitStages:
    - "RequestReceived"

The rule applies to events that do not match any previous rule in the policy file and are not at the RequestReceived stage. The rule says that get, list, and watch requests on any resource that belongs to one of the listed groups should be logged at the Request level.

The next rule in the policy file looks like this:

- level: RequestResponse
  resources:
    - group: "" # core
    - group: "admissionregistration.k8s.io"
    - group: "apiextensions.k8s.io"
    - group: "apiregistration.k8s.io"
    - group: "apps"
    - group: "authentication.k8s.io"
    - group: "authorization.k8s.io"
    - group: "autoscaling"
    - group: "batch"
    - group: "certificates.k8s.io"
    - group: "extensions"
    - group: "metrics.k8s.io"
    - group: "networking.k8s.io"
    - group: "policy"
    - group: "rbac.authorization.k8s.io"
    - group: "settings.k8s.io"
    - group: "storage.k8s.io"
  omitStages:
    - "RequestReceived"

The rule applies to events that do not match any previous rule in the policy file and are not at the RequestReceived stage. In particular, the rule does not apply to the read requests: get, list, and watch. Instead, the rule applies to write requests like create, update, and delete. The rule says that write requests should be logged at the RequestResponse level.

What's next