Skip to content

Latest commit

 

History

History
257 lines (181 loc) · 9.61 KB

README.md

File metadata and controls

257 lines (181 loc) · 9.61 KB
Raw

Integrate With Loki

Overview

Category
Signal types logs
Backend type custom in-cluster
OTLP-native no

Learn how to use Loki as a logging backend with Kyma's LogPipeline or with Promtail.

Warning

This guide uses Grafana Loki, which is distributed under AGPL-3.0 only. Using components that have this license might affect the license of your project. Inform yourself about the license used by Grafana Loki under https://grafana.com/licensing/).

setup

Table of Content

Prerequisites

Preparation

  1. Export your namespace as a variable with the following command:

    export K8S_NAMESPACE="loki"

  2. Export the Helm release names that you want to use. It can be any name, but be aware that all resources in the cluster will be prefixed with that name. Run the following command:

    export HELM_LOKI_RELEASE="loki"

  3. Update your Helm installation with the required Helm repository:

    helm repo add grafana https://grafana.github.io/helm-charts
helm repo update

Loki Installation

Depending on your scalability needs and storage requirements, you can install Loki in different Deployment modes. The following instructions install Loki in a lightweight in-cluster solution that does not fulfil production-grade qualities. Consider using a scalable setup based on an object storage backend instead (see Install the simple scalable Helm chart).

Install Loki

You install the Loki stack with a Helm upgrade command, which installs the chart if not present yet.

helm upgrade --install --create-namespace -n ${K8S_NAMESPACE} ${HELM_LOKI_RELEASE} grafana/loki \
  -f https://raw.githubusercontent.com/grafana/loki/main/production/helm/loki/single-binary-values.yaml \
  --set-string 'loki.podLabels.sidecar\.istio\.io/inject=true' \
  --set 'singleBinary.resources.requests.cpu=1' \
  --set 'loki.auth_enabled=false'

The previous command uses an example values.yaml from the Loki repository for setting up Loki in the 'SingleBinary' mode. Additionally, it applies:

  • Istio sidecar injection for the Loki instance
  • a reduced CPU request setting for smaller cluster setups
  • disabled multi-tenancy for easier setup

Alternatively, you can create your own values.yaml file and adjust the command.

Verify Loki Installation

Check that the loki Pod has been created in the Namespace and is in the Running state:

kubectl -n ${K8S_NAMESPACE} get pod -l app.kubernetes.io/name=loki

Log Agent Installation

Install the Log Agent

To ingest the application logs from within your cluster to Loki, you can either choose an installation based on Promtail, which is the log collector recommended by Loki and provides a ready-to-use setup. Alternatively, you can use Kyma's LogPipeline feature based on Fluent Bit.

Install Promtail

To install Promtail pointing it to the previously installed Loki instance, run:

helm upgrade --install --create-namespace -n ${K8S_NAMESPACE} promtail grafana/promtail -f https://raw.githubusercontent.com/kyma-project/telemetry-manager/main/docs/user/integration/loki/promtail-values.yaml --set "config.clients[0].url=https://${HELM_LOKI_RELEASE}.${K8S_NAMESPACE}.svc.cluster.local:3100/loki/api/v1/push"

Install Fluent Bit with Kyma's LogPipeline

Warning

This setup uses an unsupported output plugin for the LogPipeline.

Apply the LogPipeline:

cat <<EOF | kubectl apply -f -
apiVersion: telemetry.kyma-project.io/v1alpha1
kind: LogPipeline
metadata:
  name: custom-loki
spec:
   input:
      application:
         namespaces:
           system: true
   output:
      custom: |
         name   loki
         host   ${HELM_LOKI_RELEASE}.${K8S_NAMESPACE}.svc.cluster.local
         port   3100
         auto_kubernetes_labels off
         labels job=fluentbit, container=\$kubernetes['container_name'], namespace=\$kubernetes['namespace_name'], pod=\$kubernetes['pod_name'], node=\$kubernetes['host'], app=\$kubernetes['labels']['app'],app=\$kubernetes['labels']['app.kubernetes.io/name']
EOF

When the status of the applied LogPipeline resource turns into Running, the underlying Fluent Bit is reconfigured and log shipment to your Loki instance is active.

Note

The used output plugin configuration uses a static label map to assign labels of a Pod to Loki log streams. It's not recommended to activate the auto_kubernetes_labels feature for using all labels of a Pod because this lowers the performance. Follow Loki's labelling best practices for a tailor-made setup that fits your workload configuration.

Verify the Setup by Accessing Logs Using the Loki API

  1. To access the Loki API, use kubectl port forwarding. Run:

    kubectl -n ${K8S_NAMESPACE} port-forward svc/$(kubectl  get svc -n ${K8S_NAMESPACE} -l app.kubernetes.io/name=loki -ojsonpath='{.items[0].metadata.name}') 3100

  2. Loki queries need a query parameter time, provided in nanoseconds. To get the current nanoseconds in Linux or macOS, run:

    date +%s

  3. To get the latest logs from Loki, replace the {NANOSECONDS} placeholder with the result of the previous command, and run:

    curl -G -s  "http://localhost:3100/loki/api/v1/query" \
  --data-urlencode \
  'query={job="fluentbit"}' \
  --data-urlencode \
  'time={NANOSECONDS}'

Grafana Installation

Because Grafana provides a very good Loki integration, you might want to install it as well.

Install Grafana

  1. To deploy Grafana, run:

    helm upgrade --install --create-namespace -n ${K8S_NAMESPACE} grafana grafana/grafana -f https://raw.githubusercontent.com/kyma-project/telemetry-manager/main/docs/user/integration/loki/grafana-values.yaml

  2. To enable Loki as Grafana data source, run:

    cat <<EOF | kubectl apply -n ${K8S_NAMESPACE} -f -
apiVersion: v1
kind: ConfigMap
metadata:
  name: grafana-loki-datasource
  labels:
    grafana_datasource: "1"
data:
   loki-datasource.yaml: |-
      apiVersion: 1
      datasources:
      - name: Loki
        type: loki
        access: proxy
        url: "http://${HELM_LOKI_RELEASE}:3100"
        version: 1
        isDefault: false
        jsonData: {}
EOF

  3. Get the password to access the Grafana UI:

    kubectl get secret --namespace ${K8S_NAMESPACE} grafana -o jsonpath="{.data.admin-password}" | base64 --decode ; echo

  4. To access the Grafana UI with kubectl port forwarding, run:

    kubectl -n ${K8S_NAMESPACE} port-forward svc/grafana 3000:80

  5. In your browser, open Grafana under http://localhost:3000 and log in with user admin and the password you retrieved before.

Grafana Exposure

Expose Grafana

  1. To expose Grafana using the Kyma API Gateway, create an APIRule:

    kubectl -n ${K8S_NAMESPACE} apply -f https://raw.githubusercontent.com/kyma-project/telemetry-manager/main/docs/user/integration/loki/apirule.yaml

  2. Get the public URL of your Loki instance:

    kubectl -n ${K8S_NAMESPACE} get virtualservice -l apirule.gateway.kyma-project.io/v1beta1=grafana.${K8S_NAMESPACE} -ojsonpath='{.items[*].spec.hosts[*]}'

Add a Link for Grafana to the Kyma Dashboard

  1. Download the kyma-dashboard-configmap.yaml file and change {GRAFANA_LINK} to the public URL of your Grafana instance.

    curl https://raw.githubusercontent.com/kyma-project/telemetry-manager/main/docs/user/integration/loki/kyma-dashboard-configmap.yaml -o kyma-dashboard-configmap.yaml

  2. Optionally, adjust the ConfigMap: You can change the label field to change the name of the tab. If you want to move it to another category, change the category tab.

  3. Apply the ConfigMap, and go to Kyma dashboard. Under the Observability section, you should see a link to the newly exposed Grafana. If you already have a busola-config, merge it with the existing one:

    kubectl apply -f dashboard-configmap.yaml

Clean Up

  1. To remove the installation from the cluster, run:

    helm delete -n ${K8S_NAMESPACE} ${HELM_LOKI_RELEASE}
helm delete -n ${K8S_NAMESPACE} promtail
helm delete -n ${K8S_NAMESPACE} grafana

  2. To remove the deployed LogPipeline instance from cluster, run:

    kubectl delete LogPipeline custom-loki