Enabling optional features on the in-cluster control plane

This page describes how to enable optional features on an in-cluster control plane. For information on the managed control plane, see Configure managed Anthos Service Mesh.

When you install Anthos Service Mesh, the features that are enabled by default differ by platform. You enable optional features by including an overlay file when you install (or upgrade) Anthos Service Mesh. An overlay file is a YAML file containing an IstioOperator custom resource (CR) that you use to configure the control plane. You can override the default configuration and enable an optional feature or disable a default feature in an overlay file. Specify one feature per overlay file. You can layer on more overlays, and each overlay file overrides the configuration on the previous layers.

Caution: Using asmcli install overwrites the existing control plane configuration. When you enable optional features to an existing installation, make sure to specify all files that you want to keep in the original configuration, as well as the overlay files for the features that you want to enable. We recommended that you store the overlay files that you use for installation in source control so that you can apply the same configurations to your control plane with upgrades or when enabling additional optional features.

About the overlay files

The overlay files on this page are in the anthos-service-mesh package in GitHub. These files contain common customizations to the default configuration. You can use these files as they are, or you can make additional changes to them as needed.

When you install Anthos Service Mesh using the Google-provided asmcli script, you can specify one or more overlay files with the --option or the --custom_overlay options. If you don't need to make any changes to the files in the anthos-service-mesh repository, you can use --option, and the script fetches the file from GitHub for you. Otherwise, you can make changes to the overlay file, and then use the --custom_overlay option to pass it to the asmcli.

Important: Don't specify more than one IstioOperator CR in an overlay file because istioctl install only applies the first CR that it finds in the file.

Don't include multiple CRs in one overlay file	Create separate overlay files for each CR

How to enable optional features

The following examples are simplified to show only using the custom overlays to enable optional features. Replace OTHER_FLAGS with the other command line options.

Important: The order of the files on the command line matters. Be sure to specify the configuration required for the default features first, and then the overlay files after that.

The asmcli install command provides two ways to enable an optional feature. The method that you use depends on whether you need to make changes to the overlay file.

• Use --option when you don't need to make any changes to the overlay file. With --option, asmcli fetches the file from the GitHub repository for you, so you must have an internet connection.

  ./asmcli install \
    OTHER_FLAGS \
    --option OPTION_NAME
  

  Replace OPTION_NAME with the option you want to enable. For a list of options refer to the anthos-service-mesh package.

• Use --custom_overlay when you need to customize the overlay file.

  ./asmcli install \
    OTHER_FLAGS \
    --custom_overlay PATH_TO_FILE
  

  Replace PATH_TO_FILE with the path to the overlay file you want to use.

YAML for optional features

The following sections provide the YAML to enable optional and supported features.

mTLS STRICT mode

The global.mtls.enabled configuration was removed from the IstioOperator CR to avoid issues with upgrades and to provide a more flexible installation. To enable STRICT mTLS, configure a peer authentication policy instead.

Distroless proxy image

As a best practice, you should restrict the contents of a container runtime to only the necessary packages. This approach improves security and the signal-to-noise ratio of Common Vulnerabilities and Exposures (CVE) scanners. Istio provides proxy images based on distroless base images.

The following configuration enables distroless images for the entire Anthos Service Mesh. An image type change requires each pod to restart and get re-injected to take effect.

apiVersion: install.istio.io/v1alpha1
kind: IstioOperator
spec:
  meshConfig:
    defaultConfig:
      image:
        imageType: distroless

The distroless image does not contain any binaries other than the proxy. It is therefore not possible to exec a shell or use curl, ping, or other debug utilities inside the container. If you try to execute a shell you will see the following error.

error: Internal error occurred: error executing command in container: failed to exec in container: failed to start exec "<container-id>"
OCI runtime exec failed: exec failed: container_linux.go:380: starting container process caused: exec: "sh": executable file not found in $PATH: unknown

If you need access to these tools for specific pods, you may override the imageType by using the following pod annotation.

sidecar.istio.io/proxyImageType: debug

After changing the image type of a deployment via the annotation, the deployment should be restarted.

kubectl rollout restart deployment -n NAMESPACE DEPLOYMENT_NAME

For most types of proxy debugging, istioctl proxy-cmd should be used which does not require a debug base image.

Direct Envoy to stdout

---
apiVersion: install.istio.io/v1alpha1
kind: IstioOperator
spec:
  meshConfig:
    accessLogFile: "/dev/stdout"

For more information, see Enable Envoy's access logging.

Cloud Trace

Cloud Trace is available with Anthos Service Mesh installations on the following platforms:

• GKE on Google Cloud
• GKE Enterprise clusters on-premises if you install with Anthos Service Mesh certificate authority (Mesh CA)

For detailed pricing information, refer to the Cloud Trace pricing page.

---
apiVersion: install.istio.io/v1alpha1
kind: IstioOperator
spec:
  meshConfig:
    enableTracing: true
  values:
    global:
      proxy:
        tracer: stackdriver

The default sampling rate is 1%, but you can override the default by specifying a tracing.sampling value. The value must be in the range of 0.0 to 100.0 with a precision of 0.01. For example, to trace 5 requests out of every 10,000, use 0.05.

The following example shows a sampling rate of 100% (which you would only do for demo or troubleshooting purposes).

apiVersion: install.istio.io/v1alpha1
kind: IstioOperator
spec:
  meshConfig:
    enableTracing: true
    defaultConfig:
      tracing:
        sampling: 100
  values:
    global:
      proxy:
        tracer: stackdriver

Note that currently tracer configuration is part of proxy bootstrap configuration, so pod needs to restart and get re-injected to pick up the tracer update. For example, you can use the following command the restart pods belong to a deployment:

kubectl rollout restart deployment -n NAMESPACE DEPLOYMENT_NAME

Trace context propagation

Although the sidecar proxies can automatically send trace spans, they need some hints to tie together the entire trace. Applications need to propagate the appropriate HTTP headers so that when the proxies send span information, the spans can be correlated correctly into a single trace.

To do this, an application needs to collect and propagate the appropriate headers from the incoming request to any outgoing requests. The Anthos Service Mesh Stackdriver tracing configuration will accept any of the following header formats, and will propagate all of the following formats:

• B3 (x-b3-traceid, x-b3-spanid, x-b3parentspanid, x-b3-sampled, x-b3-flags)
• W3C TraceContext (traceparent)
• Google Cloud Trace (x-cloud-trace-context)
• gRPC TraceBin (grpc-trace-bin)

This means that your applications can use any of those formats to propagate tracing context and the traces will be generated and set to Stackdriver appropriately.

Example

Here is an example HTTP-Get request with a traceparent header in the original request. Notice the additional trace context headers added by the proxy.

$ kubectl exec -it sleep-557747455f-n6flv -- curl "httpbin:8000/anything?freeform=" -H "accept: application/json" -H "Traceparent: 00-7543d15e09e5d61801d4f74cde1269b8-604ef051d35c5b3f-01" -vv
*   Trying 10.12.3.52:8000...
* Connected to httpbin (10.12.3.52) port 8000 (#0)
> GET /anything?freeform= HTTP/1.1
> Host: httpbin:8000
> User-Agent: curl/7.80.0-DEV
> accept: application/json
> Traceparent: 00-7543d15e09e5d61801d4f74cde1269b8-604ef051d35c5b3f-01
>
* Mark bundle as not supporting multiuse
< HTTP/1.1 200 OK
< server: envoy
< date: Wed, 10 Nov 2021 20:36:04 GMT
< content-type: application/json
< content-length: 1032
< access-control-allow-origin: *
< access-control-allow-credentials: true
< x-envoy-upstream-service-time: 5
<
{
  "args": {
    "freeform": ""
  },
  "data": "",
  "files": {},
  "form": {},
  "headers": {
    "Accept": "application/json",
    "Grpc-Trace-Bin": "AAB1Q9FeCeXWGAHU90zeEmm4AaDHmGRtdM7wAgE",
    "Host": "httpbin:8000",
    "Traceparent": "00-7543d15e09e5d61801d4f74cde1269b8-a0c798646d74cef0-01",
    "User-Agent": "curl/7.80.0-DEV",
    "X-B3-Sampled": "1",
    "X-B3-Spanid": "a0c798646d74cef0",
    "X-B3-Traceid": "7543d15e09e5d61801d4f74cde1269b8",
    "X-Cloud-Trace-Context": "7543d15e09e5d61801d4f74cde1269b8/11585396123534413552;o=1",
    "X-Envoy-Attempt-Count": "1",
    "X-Forwarded-Client-Cert": "<REDACTED>"
  },
  "json": null,
  "method": "GET",
  "origin": "127.0.0.6",
  "url": "http://httpbin:8000/anything?freeform="
}

Notice that in the returned set of request headers, the full set of trace context headers is present.

For more examples propagating the headers, see Trace context propagation.

Create a trace from client with custom ID

To create a trace from a client with a custom ID, use the curl command to create a request with an external client and force it to show a trace. For example:

curl $URL --header "x-client-trace-id: 105445aa7843bc8bf206b12000100000"

For more information about x-client-trace-id, refer to the Envoy documentation.

Note: If you are testing the Anthos Service Mesh sidecar containers sending Cloud Trace data using the Google microservices-demo, make sure that you disable the built-in Cloud Trace functionality. The microservices-demo is able to send Cloud Trace data without Istio. Not disabling this functionality will result in duplicated traces because both the application container and the sidecar container will send traces. There are several DISABLE_TRACING variables to configure in the microservices-demo release file.

Egress via egress gateways

We recommend that you install an injected gateway as described in Install and upgrade gateways.

Istio Container Network Interface

How you enable the Istio Container Network Interface (CNI) depends on the environment that Anthos Service Mesh is installed on.

1. Enable a network policy.

2. Choose the overlay file that matches your platform.

Enable CNI on GKE

---
apiVersion: install.istio.io/v1alpha1
kind: IstioOperator
spec:
  components:
    cni:
      enabled: true
      namespace: kube-system
  values:
    cni:
      cniBinDir: /home/kubernetes/bin
      excludeNamespaces:
        - istio-system
        - kube-system

Enable CNI on-premises

---
apiVersion: install.istio.io/v1alpha1
kind: IstioOperator
spec:
  components:
    cni:
      enabled: true
      namespace: kube-system
  values:
    cni:
      cniBinDir: /opt/cni/bin
      excludeNamespaces:
        - istio-system
        - kube-system
        - gke-system

Enable Google Cloud Observability for off-Google Cloud

Installing Anthos Service Mesh with Istio CA outside of Google Cloud reports metrics to Prometheus by default. Use this option to enable reporting metrics to Google Cloud Observability instead, or both Prometheus and Stackdriver, so you can use the Anthos Service Mesh dashboards.

Only Stackdriver

---
apiVersion: install.istio.io/v1alpha1
kind: IstioOperator
spec:
  values:
    telemetry:
      enabled: true
      v2:
        enabled: true
        prometheus:
          enabled: false
        stackdriver:
          enabled: true

Stackdriver and Prometheus

---
apiVersion: install.istio.io/v1alpha1
kind: IstioOperator
spec:
  values:
    telemetry:
      enabled: true
      v2:
        enabled: true
        prometheus:
          enabled: true
        stackdriver:
          enabled: true

Enable an internal load balancer

We recommend that you install an injected gateway as described in Install and upgrade gateways to set up an internal load balancer on GKE. When configuring the gateway Service, you include the annotation: networking.gke.io/load-balancer-type: "Internal"

External certificate management on the ingress gateway

For information on enabling external certificate management on the ingress gateway using Envoy SDS, see Secure Gateways.