Installing logging, metrics, and traces

If you followed one of the comprehensive install guides or you performed a custom installation and included the monitoring.yaml file in your installation, all of the observability features are already installed and you can skip down to the Create Elasticsearch Indices section.

If you have not yet installed any observability plugins, continue to the next sections to do so now.

Metrics

  1. Run the following command and follow the instructions below to enable request metrics if they are wanted:

    kubectl edit cm -n knative-serving config-observability
    

Add metrics.request-metrics-backend-destination: prometheus to data field. You can find detailed information in data._example field in the ConfigMap you are editing.

  1. Run the following command to install Prometheus and Grafana:
   kubectl apply --filename https://github.com/knative/serving/releases/download/v0.8.0/monitoring-metrics-prometheus.yaml
  1. Ensure that the grafana-*, kibana-logging-*, kube-state-metrics-*, node-exporter-* and prometheus-system-* pods all report a Running status:
   kubectl get pods --namespace knative-monitoring --watch

For example:

   NAME                                  READY     STATUS    RESTARTS   AGE
   grafana-798cf569ff-v4q74              1/1       Running   0          2d
   kibana-logging-7d474fbb45-6qb8x       1/1       Running   0          2d
   kube-state-metrics-75bd4f5b8b-8t2h2   4/4       Running   0          2d
   node-exporter-cr6bh                   2/2       Running   0          2d
   node-exporter-mf6k7                   2/2       Running   0          2d
   node-exporter-rhzr7                   2/2       Running   0          2d
   prometheus-system-0                   1/1       Running   0          2d
   prometheus-system-1                   1/1       Running   0          2d

Tip: Hit CTRL+C to exit watch mode.

Accessing Metrics for more information about metrics in Knative.

Logs

Enable Request Logs

Run the following command and follow the instructions below to enable request logs if they are wanted:

kubectl edit cm -n knative-serving config-observability

Copy logging.request-log-template from data._example field todata field in the ConfigMap you are editing.. You can find detailed information in data._example field to customize the request log format.

Choose One Logging Backend

Knative offers three different setups for collecting logs. Choose one to install:

  1. Elasticsearch and Kibana
  2. Stackdriver
  3. Custom logging plugin

Elasticsearch and Kibana

  1. Run the following command to install an ELK stack:
   kubectl apply --filename https://github.com/knative/serving/releases/download/v0.8.0/monitoring-logs-elasticsearch.yaml
  1. Ensure that the elasticsearch-logging-*, fluentd-ds-*, and kibana-logging-* pods all report a Running status:
   kubectl get pods --namespace knative-monitoring --watch

For example:

   NAME                                  READY     STATUS    RESTARTS   AGE
   elasticsearch-logging-0               1/1       Running   0          2d
   elasticsearch-logging-1               1/1       Running   0          2d
   fluentd-ds-5kc85                      1/1       Running   0          2d
   fluentd-ds-vhrcq                      1/1       Running   0          2d
   fluentd-ds-xghk9                      1/1       Running   0          2d
   kibana-logging-7d474fbb45-6qb8x       1/1       Running   0          2d

Tip: Hit CTRL+C to exit watch mode.

  1. Verify that each of your nodes have the beta.kubernetes.io/fluentd-ds-ready=true label:
   kubectl get nodes --selector beta.kubernetes.io/fluentd-ds-ready=true
  1. If you receive the No Resources Found response:

    1. Run the following command to ensure that the Fluentd DaemonSet runs on all your nodes:
      kubectl label nodes --all beta.kubernetes.io/fluentd-ds-ready="true"
    1. Run the following command to ensure that the fluentd-ds daemonset is ready on at least one node:
      kubectl get daemonset fluentd-ds --namespace knative-monitoring --watch

    Tip: Hit CTRL+C to exit watch mode.

  2. When the installation is complete and all the resources are running, you can continue to the next section and begin creating your Elasticsearch indices.

Create Elasticsearch Indices

To visualize logs with Kibana, you need to set which Elasticsearch indices to explore.

  • To open the Kibana UI (the visualization tool for Elasticsearch), you must start a local proxy by running the following command:
  kubectl proxy

This command starts a local proxy of Kibana on port 8001. For security reasons, the Kibana UI is exposed only within the cluster.

  • Navigate to the Kibana UI. It might take a couple of minutes for the proxy to work.

  • Within the “Configure an index pattern” page, enter logstash-* to Index pattern and select @timestamp from Time Filter field name and click on Create button.

Create logstash-* index

See Accessing Logs for more information about logs in Knative.

Stackdriver

To configure and setup monitoring:

  1. Clone the Knative Serving repository:

    git clone -b v0.8.0 https://github.com/knative/serving knative-serving
    cd knative-serving
    1. Choose a container image that meets the Fluentd image requirements. For example, you can use a public image. Or you can create a custom one and upload the image to a container registry which your cluster has read access to.

    You must configure and build your own Fluentd image if either of the following are true:

    • Your Knative Serving component is not hosted on a Google Cloud Platform (GCP) based cluster.
    • You want to send logs to another GCP project.
    1. Follow the instructions in “Setting up a logging plugin” to configure the stackdriver components settings.

    2. Install Knative Stackdriver components by running the following command from the root directory of knative/serving repository:

      kubectl apply --recursive --filename config/monitoring/100-namespace.yaml \
          --filename config/monitoring/logging/stackdriver
  2. Ensure that the fluentd-ds-* pods all report a Running status:

    kubectl get pods --namespace knative-monitoring --watch

    For example:

    NAME                                  READY     STATUS    RESTARTS   AGE
    fluentd-ds-5kc85                      1/1       Running   0          2d
    fluentd-ds-vhrcq                      1/1       Running   0          2d
    fluentd-ds-xghk9                      1/1       Running   0          2d

    Tip: Hit CTRL+C to exit watch mode.

  3. Verify that each of your nodes have the beta.kubernetes.io/fluentd-ds-ready=true label:

    kubectl get nodes --selector beta.kubernetes.io/fluentd-ds-ready=true
    1. If you receive the No Resources Found response:

    2. Run the following command to ensure that the Fluentd DaemonSet runs on all your nodes:

      kubectl label nodes --all beta.kubernetes.io/fluentd-ds-ready="true"
      1. Run the following command to ensure that the fluentd-ds daemonset is ready on at least one node:
      kubectl get daemonset fluentd-ds --namespace knative-monitoring

    See Accessing Logs for more information about logs in Knative.

    End to end request tracing

    You can choose from one of the following options to enable request tracing in your Knative Serving cluster.

    Important: Your cluster supports only a single request trace tool. If you want to replace a currently installed request trace tool, you must first uninstall that tool before installing the new tool.

    Zipkin

    1. Install support for Zipkin:
    • If Elasticsearch is not installed or if you don’t want to persist end to end traces, run:
     kubectl apply --filename https://github.com/knative/serving/releases/download/v0.8.0/monitoring-tracing-zipkin-in-mem.yaml
    • If Elasticsearch is installed and you want to persist end to end traces, first run:
     kubectl apply --filename https://github.com/knative/serving/releases/download/v0.8.0/monitoring-tracing-zipkin.yaml
    1. Create an Elasticsearch index for end to end traces:

    2. Open Kibana UI as described in Create Elasticsearch Indices section.

    3. Select Create Index Pattern button on top left of the page. Enter zipkin* to Index pattern and select timestamp_millis from Time Filter field name and click on Create button.

    Visit Accessing Traces for more information on end to end traces.

    Jaeger

    1. Install the Jaeger operator. Use the instructions in jaegertracing/jaeger-operator repository and follow only the steps in the Installing the operator section.

    2. Install support for Jaeger:

    • If Elasticsearch is not installed or if you don’t want to persist end to end traces, run:
     kubectl apply --filename https://github.com/knative/serving/releases/download/v0.8.0/monitoring-tracing-jaeger-in-mem.yaml
    • If Elasticsearch is installed and you want to persist end to end traces, first run:
     kubectl apply --filename https://github.com/knative/serving/releases/download/v0.8.0/monitoring-tracing-jaeger.yaml
    1. Create an Elasticsearch index for end to end traces:

    2. Open Kibana UI as described in Create Elasticsearch Indices section.

    3. Select Create Index Pattern button on top left of the page. Enter jaeger* to Index pattern and select timestamp_millis from Time Filter field name and click on Create button.

    Visit Accessing Traces for more information on end to end traces.

    Learn More