Enable application logging and monitoring

This page shows how to configure a cluster for GKE on Bare Metal so that custom logs and metrics from user applications are sent to Cloud Logging and Cloud Monitoring and Google Cloud Managed Service for Prometheus.

Enable Managed Service for Prometheus by setting enableGMPForApplications to true in the Stackdriver object. This configuration lets you monitor and alert on your workloads globally, using Prometheus. For instructions and additional information, see Enable Managed Service for Prometheus on this page.
Enable Cloud Logging for user applications by setting enableCloudLoggingForApplications to true in the Stackdriver object. This configuration provides logging for your workloads. For instructions and additional information, see Enable Cloud Logging for user applications on this page.

Enable Managed Service for Prometheus

The configuration for Managed Service for Prometheus is specified in a Stackdriver object named stackdriver. For additional information, including best practices and troubleshooting, see the Managed Service for Prometheus documentation.

To configure the stackdriver object to enable Google Cloud Managed Service for Prometheus:

Open the stackdriver object for editing:
```
kubectl --kubeconfig=CLUSTER_KUBECONFIG \
    --namespace kube-system edit stackdriver stackdriver
```
Replace CLUSTER_KUBECONFIG with the path of your cluster kubeconfig file.

Under spec, set enableGMPForApplications to true:

apiVersion: addons.gke.io/v1alpha1
kind: Stackdriver
metadata:
  name: stackdriver
  namespace: kube-system
spec:
  projectID: ...
  clusterName: ...
  clusterLocation: ...
  proxyConfigSecretName: ...
  enableGMPForApplications: true
  enableVPC: ...
  optimizedMetrics: true

Save and close the edited file.

The Google-managed Prometheus components start automatically in the cluster in the gmp-system namespace.

Check the Google-managed Prometheus components:

kubectl --kubeconfig=CLUSTER_KUBECONFIG --namespace gmp-system get pods

The output of this command is similar to the following:

NAME                              READY   STATUS    RESTARTS        AGE
collector-abcde                   2/2     Running   1 (5d18h ago)   5d18h
collector-fghij                   2/2     Running   1 (5d18h ago)   5d18h
collector-klmno                   2/2     Running   1 (5d18h ago)   5d18h
gmp-operator-68d49656fc-abcde     1/1     Running   0               5d18h
rule-evaluator-7c686485fc-fghij   2/2     Running   1 (5d18h ago)   5d18h

Managed Service for Prometheus supports rule evaluation and alerting. To set up rule evaluation, see Rule evaluation.

Run an example application

The managed service provides a manifest for an example application, prom-example, that emits Prometheus metrics on its metrics port. The application uses three replicas.

To deploy the application:

Create the gmp-test namespace for resources that you create as part of the example application:
```
kubectl --kubeconfig=CLUSTER_KUBECONFIG create ns gmp-test
```

Apply the application manifest with the following command:

kubectl -n gmp-test apply \
    -f https://raw.githubusercontent.com/GoogleCloudPlatform/prometheus-engine/v0.4.1/examples/example-app.yaml

Configure a PodMonitoring resource

In this section, you configure a PodMonitoring custom resource to capture metrics data emitted by the example application and send it to Managed Service for Prometheus. The PodMonitoring custom resource uses target scraping. In this case, the collector agents scrape the /metrics endpoint to which the sample application emits data.

A PodMonitoring custom resource scrapes targets in the namespace in which it's deployed only. To scrape targets in multiple namespaces, deploy the same PodMonitoring custom resource in each namespace. You can verify the PodMonitoring resource is installed in the intended namespace by running the following command:

kubectl --kubeconfig CLUSTER_KUBECONFIG get podmonitoring -A

For reference documentation about all the Managed Service for Prometheus custom resources, see the prometheus-engine/doc/api reference.

The following manifest defines a PodMonitoring resource, prom-example, in the gmp-test namespace. The resource finds all Pods in the namespace that have the label app with the value prom-example. The matching Pods are scraped on a port named metrics, every 30 seconds, on the /metrics HTTP path.

apiVersion: monitoring.googleapis.com/v1
kind: PodMonitoring
metadata:
  name: prom-example
spec:
  selector:
    matchLabels:
      app: prom-example
  endpoints:
  - port: metrics
    interval: 30s

To apply this resource, run the following command:

kubectl --kubeconfig CLUSTER_KUBECONFIG -n gmp-test apply \
    -f https://raw.githubusercontent.com/GoogleCloudPlatform/prometheus-engine/v0.4.1/examples/pod-monitoring.yaml

Managed Service for Prometheus is now scraping the matching Pods.

Query metrics data

The simplest way to verify that your Prometheus data is being exported is to use PromQL queries in the Metrics Explorer in the Google Cloud console.

To run a PromQL query, do the following:

In the Google Cloud console, go to the Monitoring page or click the following button:

Go to Monitoring
In the navigation pane, select Metrics Explorer.
Use Prometheus Query Language (PromQL) to specify the data to display on the chart:
1. In the toolbar of the Select a metric pane, select Code Editor.
2. Select PromQL in the Language toggle. The language toggle is at the bottom of the Code Editor pane.
3. Enter your query into the query editor. For example, to chart the average number of seconds CPUs spent in each mode over the past hour, use the following query:
```
avg(rate(kubernetes_io:anthos_container_cpu_usage_seconds_total
{monitored_resource="k8s_node"}[1h]))
```
For more information about using PromQL, see PromQL in Cloud Monitoring.

The following screenshot shows a chart that displays the anthos_container_cpu_usage_seconds_total metric:

Managed Service for Prometheus chart for the Prometheus `anthos_container_cpu_usage_seconds_total` metric.

If you collect large amounts of data, you might want to filter exported metrics to keep down costs.

Enable Cloud Logging for user applications

The configuration for Cloud Logging and Cloud Monitoring is held in a Stackdriver object named stackdriver.

Open the stackdriver object for editing:
```
kubectl --kubeconfig=CLUSTER_KUBECONFIG \
    --namespace kube-system edit stackdriver stackdriver
```
Replace CLUSTER_KUBECONFIG with the path of your user cluster kubeconfig file.

In the spec section, set enableCloudLoggingForApplications to true:

apiVersion: addons.gke.io/v1alpha1
  kind: Stackdriver
  metadata:
    name: stackdriver
    namespace: kube-system
  spec:
    projectID: ...
    clusterName: ...
    clusterLocation: ...
    proxyConfigSecretName: ...
    enableCloudLoggingForApplications: true
    enableVPC: ...
    optimizedMetrics: true

Save and close the edited file.

Run an example application

In this section, you create an application that writes custom logs.

Save the following Deployment manifests to a file named my-app.yaml.

apiVersion: apps/v1
kind: Deployment
metadata:
  name: "monitoring-example"
  namespace: "default"
  labels:
    app: "monitoring-example"
spec:
  replicas: 1
  selector:
    matchLabels:
      app: "monitoring-example"
  template:
    metadata:
      labels:
        app: "monitoring-example"
    spec:
      containers:
      - image: gcr.io/google-samples/prometheus-dummy-exporter:latest
        name: prometheus-example-exporter
        imagePullPolicy: Always
        command:
        - /bin/sh
        - -c
        - ./prometheus-dummy-exporter --metric-name=example_monitoring_up --metric-value=1 --port=9090
        resources:
          requests:
            cpu: 100m

Create the Deployment

kubectl --kubeconfig CLUSTER_KUBECONFIG apply -f my-app.yaml

View application logs

Console

Go to the Logs Explorer in the Google Cloud console.

Go to the Logs Explorer
Click Resource. In the ALL RESOURCE TYPES menu, select Kubernetes Container.
Under CLUSTER_NAME, select the name of your user cluster.
Under NAMESPACE_NAME, select default.
Click Add and then click Run Query.

Under Query results, you can see log entries from the monitoring-example Deployment. For example:

{
  "textPayload": "2020/11/14 01:24:24 Starting to listen on :9090\n",
  "insertId": "1oa4vhg3qfxidt",
  "resource": {
    "type": "k8s_container",
    "labels": {
      "pod_name": "monitoring-example-7685d96496-xqfsf",
      "cluster_name": ...,
      "namespace_name": "default",
      "project_id": ...,
      "location": "us-west1",
      "container_name": "prometheus-example-exporter"
    }
  },
  "timestamp": "2020-11-14T01:24:24.358600252Z",
  "labels": {
    "k8s-pod/pod-template-hash": "7685d96496",
    "k8s-pod/app": "monitoring-example"
  },
  "logName": "projects/.../logs/stdout",
  "receiveTimestamp": "2020-11-14T01:24:39.562864735Z"
}

gcloud CLI

Run this command:

gcloud logging read 'resource.labels.project_id="PROJECT_ID" AND \
    resource.type="k8s_container" AND resource.labels.namespace_name="default"'

Replace PROJECT_ID with the ID of your project.

In the output, you can see log entries from the monitoring-example Deployment. For example:

insertId: 1oa4vhg3qfxidt
labels:
  k8s-pod/app: monitoring-example
  k8s- pod/pod-template-hash: 7685d96496
logName: projects/.../logs/stdout
receiveTimestamp: '2020-11-14T01:24:39.562864735Z'
resource:
  labels:
    cluster_name: ...
    container_name: prometheus-example-exporter
    location: us-west1
    namespace_name: default
    pod_name: monitoring-example-7685d96496-xqfsf
    project_id: ...
  type: k8s_container
textPayload: |
  2020/11/14 01:24:24 Starting to listen on :9090
timestamp: '2020-11-14T01:24:24.358600252Z'

Filter application logs

Application log filtering can reduce application logging billing and network traffic from the cluster to Cloud Logging. Starting with GKE on Bare Metal release 1.15.0, when enableCloudLoggingForApplications is set to true, you can filter application logs by the following criteria:

Pod labels (podLabelSelectors)
Namespaces (namespaces)
Regular expressions for log content (contentRegexes)

GKE on Bare Metal sends only the filter results to Cloud Logging.

Define application log filters

The configuration for Logging is specified in a Stackdriver object named stackdriver.

Open the stackdriver object for editing:
```
kubectl --kubeconfig USER_CLUSTER_KUBECONFIG --namespace kube-system \
    edit stackdriver stackdriver
```
Replace USER_CLUSTER_KUBECONFIG with the path to your user cluster kubeconfig file.

Add an appLogFilter section to the spec:

  apiVersion: addons.gke.io/v1alpha1
  kind: Stackdriver
  metadata:
    name: stackdriver
    namespace: kube-system
  spec:
    enableCloudLoggingForApplications: true
    projectID: ...
    clusterName: ...
    clusterLocation: ...
    appLogFilter:
      keepLogRules:
      - namespaces:
        - prod
        ruleName: include-prod-logs
      dropLogRules:
      - podLabelSelectors:
        - disableGCPLogging=yes
        ruleName: drop-logs

Save and close the edited file.
(Optional) If you're using podLabelSelectors, restart the stackdriver-log-forwarder DaemonSet to effect your changes as soon as possible:
```
kubectl --kubeconfig USER_CLUSTER_KUBECONFIG --namespace kube-system \
    rollout restart daemonset stackdriver-log-forwarder
```
Normally, podLabelSelectors are effective after 10 minutes. Restarting the DaemonSet stackdriver-log-forwarder makes the changes take effect more quickly.

Example: Include `ERROR` or `WARN` logs in `prod` namespace only

The following example illustrates an application log filter works. You define a filter that uses a namespace (prod), a regular expression (.*(ERROR|WARN).*), and a Pod label (disableGCPLogging=yes). Then, to verify that the filter works, you run a Pod in the prod namespace to test these filter conditions.

To define and test an application log filter:

Specify an application log filter in the Stackdriver object:

In the following appLogFilter example, only ERROR or WARN logs in the prod namespace are kept. Any logs for Pods with the label disableGCPLogging=yes are dropped:

apiVersion: addons.gke.io/v1alpha1
kind: Stackdriver
metadata:
  name: stackdriver
  namespace: kube-system
spec:
  ...
  appLogFilter:
    keepLogRules:
    - namespaces:
      - prod
      contentRegexes:
      - ".*(ERROR|WARN).*"
      ruleName: include-prod-logs
    dropLogRules:
    - podLabelSelectors:
      - disableGCPLogging=yes # kubectl label pods pod disableGCPLogging=yes
      ruleName: drop-logs
...

Deploy a Pod in the prod namespace and run a script that generates ERROR and INFO log entries:

kubectl --kubeconfig USER_CLUSTER_KUBECONFIG run pod1 \
    --image gcr.io/cloud-marketplace-containers/google/debian10:latest \
    --namespace prod --restart Never --command -- \
    /bin/sh -c "while true; do echo 'ERROR is 404\\nINFO is not 404' && sleep 1; done"

The filtered logs should contain the ERROR entries only, not the INFO entries.

Add the label disableGCPLogging=yes to the Pod:
```
kubectl --kubeconfig USER_CLUSTER_KUBECONFIG label pods pod1 \
    --namespace prod disableGCPLogging=yes
```
The filtered log should no longer contain any entries for the pod1 Pod.

Application log filter API definition

The definition for the application log filter is declared within the stackdriver custom resource definition.

To get the stackdriver custom resource definition, run the following command:

kubectl --kubeconfig USER_CLUSTER_KUBECONFIG get crd stackdrivers.addons.gke.io \
    --namespace kube-system -o yaml

Enable application logging and monitoring

Enable Managed Service for Prometheus

Run an example application

Configure a PodMonitoring resource

Query metrics data

Enable Cloud Logging for user applications

Run an example application

View application logs

Console

gcloud CLI

Filter application logs

Define application log filters

Example: Include ERROR or WARN logs in prod namespace only

Application log filter API definition

Example: Include `ERROR` or `WARN` logs in `prod` namespace only