You are viewing docs for an older version of Linkerd. View the latest docs.
  • GitHub
  • Slack
  • Linkerd Forum

Authorization Policy

Linkerd’s server authorization policy allows you to control which types of traffic are allowed to meshed pods. For example, you can restrict communication to a particular service to only come from certain other services; or you can enforce that mTLS must be used on a certain port; and so on.

Adding traffic policy on your services

Note

Linkerd can only enforce policy on meshed pods, i.e. pods where the Linkerd proxy has been injected. If policy is a strict requirement, you should pair the usage of these features with HA mode, which enforces that the proxy must be present when pods start up.

By default Linkerd allows all traffic to transit the mesh, and uses a variety of mechanisms, including retries and load balancing, to ensure that requests are delivered successfully.

Sometimes, however, we want to restrict which types of traffic are allowed. Linkerd’s policy features allow you to deny traffic under certain conditions. It is configured with two basic mechanisms:

  1. A set of basic default policies, which can be set at the cluster, namespace, workload, and pod level through Kubernetes annotations.
  2. Server and ServerAuthorization CRDs that specify fine-grained policy for specific ports.

These mechanisms work in conjunction. For example, a default cluster-wide policy of deny would prohibit any traffic to any meshed pod; traffic must then be explicitly allowed through the use of Server and ServerAuthorization CRDs.

Policy annotations

The config.linkerd.io/default-inbound-policy annotation can be set at a namespace, workload, and pod level, and will determine the default traffic policy at that point in the hierarchy. Valid default policies include:

  • all-unauthenticated: inbound proxies allow all connections
  • all-authenticated: inbound proxies allow only mTLS connections from other meshed pods.
  • cluster-unauthenticated: inbound proxies allow all connections from client IPs in the cluster’s clusterNetworks (must be configured at install-time).
  • cluster-authenticated: inbound proxies allow only mTLS connections from other meshed pods from IPs in the cluster’s clusterNetworks.
  • deny inbound proxies deny all connections that are not explicitly authorized.

See the Policy reference for more default policies.

Every cluster has a default policy (by default, all-unauthenticated), set at install / upgrade time. Annotations that are present at the workload or namespace level at pod creation time can override that value to determine the default policy for that pod. Note that the default policy is fixed at proxy initialization time, and thus, after a pod is created, changing the annotation will not change the default policy for that pod.

Policy CRDs

The Server and ServerAuthorization CRDs further configure Linkerd’s policy beyond the default policies. In contrast to annotations, these CRDs can be changed dynamically and policy behavior will be updated on the fly.

A Server selects a port and a set of pods that is subject to policy. This set of pods can correspond to a single workload, or to multiple workloads (e.g. port 4191 for every pod in a namespace). Once created, a Server resource denies all traffic to that port, and traffic to that port can only be enabled by creating ServerAuthorization resources.

A ServerAuthorization defines a set of allowed traffic to a Server. A ServerAuthorization can allow traffic based on any number of things, including IP address; use of mTLS; specific mTLS identities (including wildcards, to allow for namespace selection); specific Service Accounts; and more.

See the Policy reference for more on the Server and ServerAuthorization resources.

Note

Currently, Servers can only reference ports that are defined as container ports in the pod’s manifest.

Policy rejections

Any traffic that is known to be HTTP (including HTTP/2 and gRPC) that is denied by policy will result in the proxy returning an HTTP 403. Any non-HTTP traffic will be denied at the TCP level, i.e. by refusing the connection.

Note that dynamically changing the policy may result in abrupt termination of existing TCP connections.

Examples

See emojivoto-policy.yml for an example set of policy definitions for the Emojivoto sample application.