Loki Distributed

Loki is a horizontally scalable, highly available, multi-tenant log aggregation system inspired by Prometheus. It is designed to be very cost effective and easy to operate. It does not index the contents of the logs, but rather a set of labels for each log stream.

Requirements

• Kubernetes >= 1.23.0
• Kustomize >= v3.5.3
• prometheus-operator from KFD monitoring module
• grafana from KFD monitoring module (module version >=1.15.0)
• minio-ha

Prometheus Operator is necessary since we configure a ServiceMonitor to make some metrics available from loki on prometheus

Image repository and tag

• Loki image: grafana/loki
• Loki repo: Loki on Github

Configuration

Loki Distributed is deployed in the following configuration:

• Each microservice has its own Deployment/StatefulSet
• Each Deployment has its own HPA
• Common resources set as:

  resources:
    requests:
      cpu: 100m
      memory: 128Mi
    limits:
      cpu: 500m
      memory: 1024Mi

Deployment

You can deploy Loki Distributed by running the following command in the root of the project:

kustomize build | kubectl apply -f -

This project also implements a dynamic Loki datasource that our Grafana from the monitoring stack automatically fetches and configures. To see the logs, navigate in Grafana to the explore section.

Note: These instructions are only for installing Loki as a log storage solution. For complete instructions, please refer to the main README of the Logging module.