Enable application logging and monitoring

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

For the best user application logging and monitoring experience, we strongly recommend that you use the following configuration:

• Enable Google Cloud 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.

• Disable legacy Logging and Monitoring for user applications by setting enableApplication to false in the cluster resource. Disabling this capability prevents application metrics from being collected twice. Use the steps in Enable Logging and Monitoring for user applications (Legacy), but set enableApplication to false, instead of true.

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:

1. 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.

2. 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
   

3. Save and close the edited file.

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

4. 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:

1. Create the gmp-test namespace for resources that you create as part of the example application:

   kubectl --kubeconfig=CLUSTER_KUBECONFIG create ns gmp-test
   

2. 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:

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

   Go to Monitoring

2. In the navigation pane, select Metrics Explorer.

3. 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:

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.

1. 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.

2. 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
   

3. Save and close the edited file.

Run an example application

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

1. 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
   

2. Create the Deployment

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

View application logs

Filter application logs

Application log filtering can reduce application logging billing and network traffic from the cluster to Cloud Logging. Starting with Google Distributed Cloud 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)

Google Distributed Cloud sends only the filter results to Cloud Logging.

Define application log filters

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

1. 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.

2. 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
   

3. Save and close the edited file.

4. (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:

1. 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
   ...
   

2. 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.

3. 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 Logging and Monitoring for user applications (Legacy)

We strongly recommend that you follow the configuration guidance at the beginning of this document.

The following steps still work but aren't recommended. Read this related known issue before using the following steps.

To enable Logging and Monitoring for your applications, use the spec.clusterOperations.enableApplication field in the cluster configuration file.

1. Update the cluster configuration file to set enableApplication to true:

   apiVersion: v1
   kind: Namespace
   metadata:
     name: cluster-user-basic
   ---
   apiVersion: baremetal.cluster.gke.io/v1
   kind: Cluster
   metadata:
     name: user-basic
     namespace: cluster-user-basic
   spec:
     type: user
     ...
     clusterOperations:
       projectID: project-fleet
       location: us-central1
       enableApplication: true
       ...
   

2. Use bmctl update to apply your changes:

   bmctl update cluster -c CLUSTER_NAME --admin-kubeconfig=ADMIN_KUBECONFIG
   

   Replace the following:

   • CLUSTER_NAME: the name of the cluster to upgrade.
   • ADMIN_KUBECONFIG: the path to the admin cluster kubeconfig file.

Annotate workloads

To enable the collection of custom metrics from an application, add the prometheus.io/scrape: "true" annotation to the application's Service or Pod manifest, or add the same annotation to the spec.template section in Deployment or DaemonSet manifest in order to be passed to their Pods.

Run an example application

In this section, you create an application that writes custom logs and exposes a custom metric.

1. Save the following Service and Deployment manifests to a file named my-app.yaml. Notice that the Service has the annotation prometheus.io/scrape: "true":

   kind: Service
   apiVersion: v1
   metadata:
     name: "monitoring-example"
     namespace: "default"
     annotations:
       prometheus.io/scrape: "true"
   spec:
     selector:
       app: "monitoring-example"
     ports:
       - name: http
         port: 9090
   ---
   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
   

2. Create the Deployment and the Service:

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

View application logs

View application metrics in the Google Cloud console

Your example application exposes a custom metric named example_monitoring_up. You can view the values of that metric in the Google Cloud console.

1. Go to the Metrics explorer in the Google Cloud console.

   Go to the Metrics explorer

2. For Resource type, select Kubernetes Pod or Kubernetes Container.

3. For metric, select external.googleapis.com/prometheus/example_monitoring_up.

4. In the chart, you can see that example_monitoring_up has a repeated value of 1.