Request proxy logs

Cloud Service Mesh supports two different types of access logs in Cloud Logging: Traffic logs (also known as Google Cloud Observability access logs) and Envoy access logs. This page shows you how to enable, disable, view, and interpret these logs. Note that traffic logs are enabled by default.

Enabling and disabling access logs

Managed Cloud Service Mesh

Envoy access logs

Run the following command to enable Envoy access logs and disable traffic logs:

cat <<EOF | kubectl apply -n istio-system -f -
apiVersion: telemetry.istio.io/v1alpha1
kind: Telemetry
metadata:
  name: enable-envoy-disable-sd
  namespace: istio-system
spec:
  accessLogging:
  - providers:
      - name: envoy
  - providers:
      - name: stackdriver
    disabled: true
EOF

Note that the provider name for traffic log is stackdriver.

Traffic logs

By default, traffic logs are enabled and Envoy access logs are disabled. If you previously enabled Envoy access logs and want to enable traffic logs and disable Envoy access logs, run the following command:

cat <<EOF | kubectl apply -n istio-system -f -
apiVersion: telemetry.istio.io/v1alpha1
kind: Telemetry
metadata:
  name: disable-envoy-enable-sd
  namespace: istio-system
spec:
  accessLogging:
  - providers:
      - name: envoy
    disabled: true
  - providers:
      - name: stackdriver
EOF

Both

  • To enable both Envoy access logs and traffic logs, run the following command:

    cat <<EOF | kubectl apply -n istio-system -f -
    apiVersion: telemetry.istio.io/v1alpha1
    kind: Telemetry
    metadata:
      name: enable-envoy-and-sd-access-log
      namespace: istio-system
    spec:
      accessLogging:
      - providers:
          - name: envoy
          - name: stackdriver
    EOF
    
  • To disable both Envoy access logs and traffic logs, run the following command:

    cat <<EOF | kubectl apply -n istio-system -f -
    apiVersion: telemetry.istio.io/v1alpha1
    kind: Telemetry
    metadata:
      name: disable-envoy-and-sd
      namespace: istio-system
    spec:
      accessLogging:
      - providers:
          - name: envoy
        disabled: true
      - providers:
          - name: stackdriver
        disabled: true
    EOF
    

Managed istiod

Envoy access logs

Run the following commands to enable Envoy access logging:

  1. Run the following command to add accessLogFile: /dev/stdout:

    cat <<EOF | kubectl apply -f -
    apiVersion: v1
    data:
      mesh: |-
        accessLogFile: /dev/stdout
    kind: ConfigMap
    metadata:
      name: istio-release-channel
      namespace: istio-system
    EOF
    

    where release-channel is your release channel (asm-managed, asm-managed-stable, or asm-managed-rapid).

  2. Run the following command to view the configmap:

     kubectl get configmap istio-release-channel -n istio-system -o yaml
    
  3. To verify that access logging is enabled, ensure the accessLogFile: /dev/stdout line appears in the mesh: section.

    ...
    apiVersion: v1
    data:
      mesh: |
        ....
        accessLogFile: /dev/stdout
    ...
    

Traffic logs

Traffic logs are enabled by default.

In-cluster

Envoy access logs

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

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

Traffic logs

Traffic logs are enabled by default, unless Cloud Service Mesh is installed on Google Distributed Cloud with Istio CA (previously known as Citadel).

To enable traffic logs on Google Distributed Cloud with Istio CA when installing in-cluster Cloud Service Mesh, then use the flag --option stackdriver. Alternatively, you can enable traffic logs on Google Distributed Cloud with Istio CA after installing in-cluster Cloud Service Mesh.

Viewing access logs

Envoy access logs

Command line

To view Envoy access logs in the istio-proxy log, run the following command:

kubectl logs POD_NAME -n NAMESPACE_NAME -c istio-proxy

Logs Explorer

To view Envoy access logs in the Logs Explorer:

  1. Navigate to the Logs Explorer:

    Go to the Logs Explorer

  2. Select the appropriate Google Cloud project.

  3. Run the following query:

resource.type="k8s_container" \
resource.labels.container_name="istio-proxy"
resource.labels.cluster_name="CLUSTER_NAME" \
resource.labels.namespace_name="NAMESPACE_NAME" \
resource.labels.pod_name="POD_NAME"

Traffic logs

To view traffic logs in the Logs Explorer:

  1. Navigate to the Logs Explorer:

    Go to the Logs Explorer

  2. Select the appropriate Google Cloud project.

  3. Run the following query depending on whether you are viewing client or server access logs:

    Server logs

    resource.labels.cluster_name="CLUSTER_NAME" logName="projects/PROJECT_NAME/logs/server-accesslog-stackdriver"
    

    Client logs

    resource.labels.cluster_name="CLUSTER_NAME" logName="projects/PROJECT_NAME/logs/client-accesslog-stackdriver"
    

To view traffic logs in the Cloud Service Mesh page for a Service during a specified time span, follow these steps:

  1. In Google Cloud console, go to the Cloud Service Mesh page.

    Go to the Cloud Service Mesh page

  2. Under Services, select the name of the Service you want to inspect.

  3. Go to the Metrics page.

  4. Specify a time span from the Time Span drop-down menu or set a custom span with the timeline.

  5. Under Select a filter option, click View traffic logs.

The traffic log is named as server-accesslog-stackdriver and is attached to the corresponding monitored resource (k8s_container or gce_instance) your service is using. The traffic log contains the following information:

  • HTTP request properties, such as ID, URL, size, latency, and common headers.

  • Source and destination workload information, such as name, namespace, identity, and common labels.

  • If tracing is enabled, trace information, such as sampling, trace ID, and span ID.

An example log entry looks like the following:

{
  insertId: "1awb4hug5pos2qi"
  httpRequest: {
    requestMethod: "GET"
    requestUrl: "YOUR-INGRESS/productpage"
    requestSize: "952"
    status: 200
    responseSize: "5875"
    remoteIp: "10.8.0.44:0"
    serverIp: "10.56.4.25:9080"
    latency: "1.587232023s"
    protocol: "http"
  }
  resource: {
    type: "k8s_container"
    labels: {
      location: "us-central1-a"
      project_id: "YOUR-PROJECT"
      pod_name: "productpage-v1-76589d9fdc-ptnt9"
      cluster_name: "YOUR-CLUSTER-NAME"
      container_name: "productpage"
      namespace_name: "default"
    }
  }
  timestamp: "2020-04-28T19:55:21.056759Z"
  severity: "INFO"
  labels: {
    destination_principal: "spiffe://cluster.local/ns/default/sa/bookinfo-productpage"
    response_flag: "-"
    destination_service_host: "productpage.default.svc.cluster.local"
    source_app: "istio-ingressgateway"
    service_authentication_policy: "MUTUAL_TLS"
    source_name: "istio-ingressgateway-5ff85d8dd8-mwplb"
    mesh_uid: "YOUR-MESH-UID"
    request_id: "021ce752-9001-4ac6-b6d6-3b15f5d3632"
    destination_namespace: "default"
    source_principal:  "spiffe://cluster.local/ns/istio-system/sa/istio-ingressgateway-service-account"
    destination_workload: "productpage-v1"
    destination_version: "v1"
    source_namespace: "istio-system"
    source_workload: "istio-ingressgateway"
    destination_name: "productpage-v1-76589d9fdc-ptnt9"
    destination_app: "productpage"
  }
  trace: "projects/YOUR-PROJECT/traces/d4197f59b7a43e3aeff3571bac99d536"
  receiveTimestamp: "2020-04-29T03:07:14.362416217Z"
  spanId: "43226343ca2bb2b1"
  traceSampled: true
  logName: "projects/YOUR-PROJECT/logs/server-accesslog-stackdriver"
  receiveTimestamp: "2020-04-28T19:55:32.185229100Z"
}

Interpret Cloud Service Mesh telemetry

The following sections explain how to check the status of your mesh and review the various telemetry that contain helpful details to assist your troubleshooting.

Interpret control plane metrics

Managed Cloud Service Mesh

Cloud Service Mesh with a managed Cloud Service Mesh control plane doesn't support control plane metrics.

Managed istiod

Cloud Service Mesh with a managed istiod control plane doesn't support the control plane metric inspection in this section.

In-cluster

When installing Cloud Service Mesh with the in-cluster control plane, istiod exports metrics to Google Cloud Observability for monitoring, by default. istiod prefixes these metrics with istio.io/control and give insights into the control plane state, such as number of proxies connected to each control plane instance, configuration events, pushes and validations.

Observe or troubleshoot the control plane, using the following steps.

  1. Load a sample dashboard:

    git clone https://github.com/GoogleCloudPlatform/monitoring-dashboard-samples && cd monitoring-dashboard-samples/dashboards && git checkout servicemesh
  2. Install the Cloud Service Mesh dashboard:

    gcloud monitoring dashboards create --config-from-file=dashboards/servicemesh/anthos-service-mesh-control-plane-monitoring.json
  3. Look for a dashboard named Istio Control Plane Dashboard in the list. For more information, see Viewing the installed dashboard.

For the full list of metrics available, see Exported metrics.

Diagnose configuration delays

Managed Cloud Service Mesh

Cloud Service Mesh with a managed Cloud Service Mesh control plane doesn't support diagnosing configuration delays.

Managed istiod

Cloud Service Mesh with a managed istiod control plane doesn't support diagnosing configuration delays.

In-cluster

The following steps explain how to use the pilot_proxy_convergence_time metric to diagnose a delay between a configuration change and all proxies converging.

  1. Run a shell command in a pod:

    kubectl debug --image istio/base --target istio-proxy -it $(kubectl get pod -l app=pilot -o jsonpath='{.items[0].metadata.name}' -n istio-system) -n istio-system -- curl -s
  2. Access localhost:15014 and grep for convergence in metrics:

    curl http://localhost:15014/metrics | grep convergence

Interpret traffic logs

The following information explains how to use the traffic logs to troubleshoot connection problems. Traffic logs are enabled by default.

Cloud Service Mesh exports data into traffic logs that can help you debug the following types of problems:

  • Traffic flow and failures
  • End-to-end request routing

Traffic logs are enabled by default for Cloud Service Mesh installations on Google Kubernetes Engine. You can enable traffic logs by re-running asmcli install. Use the same options that you originally installed but omit the custom overlay that disabled Stackdriver.

There are two types of traffic logs:

  • Server access logs give a server-side view of requests. They are located under server-accesslog-stackdriver, attached to the k8s_container monitored resource. Use the following URL syntax to display server-side access logs:

    https://console.cloud.google.com/logs/viewer?advancedFilter=logName="projects/PROJECT_ID/logs/server-accesslog-stackdriver"&project=PROJECT_ID
  • Client access logs give a client-side view of requests. They are located under client-accesslog-stackdriver, attached to the k8s_pod monitored resource. Use the following URL syntax to display client-side access logs:

    https://console.cloud.google.com/logs/viewer?advancedFilter=logName="projects/PROJECT_ID/logs/client-accesslog-stackdriver"&project=PROJECT_ID

Access logs contain the following information:

  • HTTP request properties, such as ID, URL, size, latency, and common headers.
  • Source and destination workload information, such as name, namespace, identity, and common labels.
  • Source and destination canonical service and revision information.
  • If tracing is enabled, the logs contain trace information, such as sampling, trace ID, and span ID.

Traffic logs may contain the following labels:

  • route_name
  • upstream_cluster
  • X-Envoy-Original-Path

This is an example log entry:

{
  "insertId": "1j84zg8g68vb62z",
  "httpRequest": {
    "requestMethod": "GET",
    "requestUrl": "http://35.235.89.201:80/productpage",
    "requestSize": "795",
    "status": 200,
    "responseSize": "7005",
    "remoteIp": "10.168.0.26:0",
    "serverIp": "10.36.3.153:9080",
    "latency": "0.229384205s",
    "protocol": "http"
  },
  "resource": {
    "type": "k8s_container",
    "labels": {
      "cluster_name": "istio-e2e22",
      "namespace_name": "istio-bookinfo-1-68819",
      "container_name": "productpage",
      "project_id": "***",
      "location": "us-west2-a",
      "pod_name": "productpage-v1-64794f5db4-8xbtf"
    }
  },
  "timestamp": "2020-08-13T21:37:42.963881Z",
  "severity": "INFO",
  "labels": {
    "protocol": "http",
    "upstream_host": "127.0.0.1:9080",
    "source_canonical_service": "istio-ingressgateway",
    "source_namespace": "istio-system",
    "x-envoy-original-path": "",
    "source_canonical_revision": "latest",
    "connection_id": "32",
    "upstream_cluster": "inbound|9080|http|productpage.istio-bookinfo-1-68819.svc.cluster.local",
    "requested_server_name": "outbound_.9080_._.productpage.istio-bookinfo-1-68819.svc.cluster.local",
    "destination_version": "v1",
    "destination_workload": "productpage-v1",
    "source_workload": "istio-ingressgateway",
    "destination_canonical_revision": "v1",
    "mesh_uid": "cluster.local",
    "source_principal": "spiffe://cluster.local/ns/istio-system/sa/istio-ingressgateway-service-account",
    "x-envoy-original-dst-host": "",
    "service_authentication_policy": "MUTUAL_TLS",
    "destination_principal": "spiffe://cluster.local/ns/istio-bookinfo-1-68819/sa/bookinfo-productpage",
    "response_flag": "-",
    "log_sampled": "false",
    "destination_service_host": "productpage.istio-bookinfo-1-68819.svc.cluster.local",
    "destination_name": "productpage-v1-64794f5db4-8xbtf",
    "destination_canonical_service": "productpage",
    "destination_namespace": "istio-bookinfo-1-68819",
    "source_name": "istio-ingressgateway-6845f6d664-lnfvp",
    "source_app": "istio-ingressgateway",
    "destination_app": "productpage",
    "request_id": "39013650-4e62-9be2-9d25-78682dd27ea4",
    "route_name": "default"
  },
  "logName": "projects/***/logs/server-accesslog-stackdriver",
  "trace": "projects/***t/traces/466d77d15753cb4d7749ba5413b5f70f",
  "receiveTimestamp": "2020-08-13T21:37:48.758673203Z",
  "spanId": "633831cb1fda4fd5",
  "traceSampled": true
}

You can use this log in various ways:

  • Integrate with Cloud Trace, which is an optional feature in Cloud Service Mesh.
  • Export traffic logs to BigQuery, where you can run queries like selecting all requests take more than 5 seconds.
  • Create log-based metrics.
  • Troubleshoot 404 and 503 errors

Troubleshoot 404 and 503 errors

The following example explains how to use this log to troubleshoot when a request fails with a 404 or 503 response code.

  1. In the client access log, search for an entry like the following:

    httpRequest: {
    requestMethod: "GET"
    requestUrl: "://IP_ADDRESS/src/Util/PHP/eval-stdin.php"
    requestSize: "2088"
    status: 404
    responseSize: "75"
    remoteIp: "10.168.0.26:34165"
    serverIp: "10.36.3.149:8080"
    latency: "0.000371440s"
    protocol: "http"
    }
  2. Navigate to the labels in the access log entry. Find the response_flag field that looks like the following:

    response_flag: "NR"

    The NR value is an acronym for NoRoute, which means no route was found for the destination or there was no matching filter chain for a downstream connection. Similarly, you can use the response_flag label to troubleshoot 503 errors also.

  3. If you see 503 errors in both the client and server access logs, check that the port names set for each service match the name of the protocol in use between them. For example, if a golang binary client connects to a golang server using HTTP, but the port is named http2, the protocol will not auto-negotiate correctly.

For more information, see response flags.

Interpret Envoy access logs

The following steps explain how to use the Envoy access logs to show traffic between both ends of a connection for troubleshooting purposes.

Envoy access logs are useful for diagnosing issues like:

  • Traffic flow and failures
  • End-to-end request routing

Envoy access logs are not enabled by default in Cloud Service Mesh and can be enabled for the clusters in a mesh.

You can troubleshoot connection or request failures by generating activity in your application that triggers an HTTP request, then inspecting the associated request in the source or destination logs.

If you trigger a request appears and it appears in the source proxy logs, it indicates that iptables traffic redirection is working correctly and the Envoy proxy is handling traffic. If you see errors in the logs, generate an Envoy configuration dump and check the Envoy cluster configuration to ensure it is correct. If you see the request but the log has no errors, check the destination proxy logs instead.

If the request appears in the destination proxy logs, it indicates that the mesh itself is working correctly. If you see an error instead, run an Envoy configuration dump and verify the correct values for the traffic port set in the listener configuration.

If the problem persists after performing the previous steps, Envoy might be unable to auto-negotiate the protocol between the sidecar and its application pod. Ensure that the Kubernetes service port name, for example http-80, matches the protocol that the application uses.

Use Logs Explorer to query logs

You can use the Logs Explorer interface to query specific Envoy access logs. For example, to query all requests that have MULTUAL_TLS enabled and use protocol grpc, append following to the server access logs query:

labels.protocol="grpc" labels.service_authentication_policy="MULTUAL_TLS"

Set an access log policy

Managed Cloud Service Mesh

To configure access logs for Cloud Service Mesh with a managed Cloud Service Mesh control plane, see Enabling access logs.

Managed istiod

To configure access logs for Cloud Service Mesh with a managed istiod control plane, see Enabling access logs.

In-cluster

To set an access log policy for Cloud Service Mesh with the in-cluster control plane:

  1. Create an IstioOperator custom overlay file that includes the applicable AccessLogPolicyConfig values for your scenario.

  2. Pass this file to asmcli using the --custom_overlay option to update the in-cluster control plane configuration. For information on running asmcli install with a custom overlay file, see Install with optional features.

View service or workload-specific information

If you have an issue with a specific service or workload rather than a mesh-wide problem, inspect the individual Envoy proxies and gather relevant information from them. To gather information about a particular workload and its proxies, you can use pilot-agent:

kubectl exec POD_NAME -n NAMESPACE_NAME -c istio-proxy -- pilot-agent request GET SCOPE

In the example, SCOPE is one of the following:

  • certs - Certificates within the Envoy instance
  • clusters - Clusters with Envoy configured
  • config_dump - Dumps the Envoy configuration
  • listeners - Listeners with Envoy configured
  • logging - View and change logging settings
  • stats - Envoy statistics
  • stats/prometheus - Envoy statistics as Prometheus records

View proxy socket states

You can directly examine the state of Envoy proxy sockets by using the following process.

  1. Display a list of established sockets, including sockets in the TIME_WAIT state, which can negatively affect scalability if their count is high:

    kubectl debug --image istio/base --target istio-proxy -it POD_NAME -n NAMESPACE_NAME -- ss -anopim
  2. Display a summary of socket statistics:

    kubectl debug --image istio/base --target istio-proxy -it POD_NAME -n NAMESPACE_NAME -- ss -s

For more information, see An Introduction to the ss Command.

istio-proxy and istio-init logs

In addition, retrieve the istio-proxy logs and review its contents for any errors that might suggest the cause of the problem:

kubectl logs POD_NAME -n NAMESPACE_NAME -c istio-proxy

You can do the same for the init container:

kubectl logs POD_NAME -n NAMESPACE_NAME -c istio-init

What's next