Skip to content

Latest commit

 

History

History
106 lines (75 loc) · 4.78 KB

loki_microservices.md

File metadata and controls

106 lines (75 loc) · 4.78 KB
Raw

Scalable loki deployment

This document aims to design the deployment of a scalable version of Loki for testing purposes. This is based on microservice mode.

This requires external storage to save chunks. For a deployment without external storage, check loki simple doc.

Hack script

A hack script is provided for convenience. It assumes you have the AWS CLI installed with credentials configured. It will create a S3 bucket and configure Loki with it.

The first argument is the bucket name, second is the AWS region. Example:

./hack/loki-microservices.sh netobserv-loki eu-west-1

For a step-by-step installation instead of this hack script, read the next sections.

Folders content

You can go to the next sections to understand the configuration of this stack or directly go to deploy section.

1 - Prerequisites

To run this stack you will need to configure:

  • a service account used for deployments and statefulSets
  • a configMap containing loki configuration used in each component

2 - Components

The following components will be generated:

By default, you will get 2 pods of each component. You can change this number updating the replicas section of each yaml file independently.

Check official architecture documentation for more details. loki architecture representation

3 - Services

Related services will be created for each components. Note that querier and ingester has alternative services with -headless suffix. These services uses clusterIP: None parameter. No load balancing will be done using these since client usually use first IP returned by the DNS record. This can be usefull to list available IPs for each component.

An additionnal service called loki-microservices-memberlist will match all the components and allow discoverability between them.

Storage configuration

Edit loki Secret and replace ACCESS_KEY_ID and SECRET_ACCESS_KEY values with your bucket credentials. These will be loaded in environment variables and can be reused in config as ${ACCESS_KEY_ID} and ${SECRET_ACCESS_KEY}.

Edit loki ConfigMap and replace $(LOKI_STORE_NAME) and $(LOKI_STORE) with your proper external storage configuration.

Example using s3 bucket called loki in us-east-1 region with credentials from secret:

  config.yaml: |
    ...
    common:
      storage:
        s3:
          s3: https://s3.us-east-1.amazonaws.com
          bucketnames: loki
          region: us-east-1
          access_key_id: ${ACCESS_KEY_ID}
          secret_access_key: ${SECRET_ACCESS_KEY}
          s3forcepathstyle: true
    ...
    schema_config:
      configs:
        - ...
          object_store: s3

Check official examples.

/!\ Without external storage, components will not share the chunks and querier will crash as soon as cache is cleared /!\

Deploy

Ensure storage configuration is up to date. Run the following command to apply all yamls at once in netobserv namespace: kubectl apply -f ./examples/loki-microservices/ -n netobserv --recursive

This will create all the components and services described above.

Destroy

Delete everything by running:

kubectl delete --recursive -f ./examples/loki-microservices

Network Observability Operator

To connect NOO with this config, you will have to update the loki section with the following urls assuming you used netobserv namespace:

  loki:
    mode: Microservices
    microservices:
      ingesterUrl: 'http://loki-microservices-distributor.netobserv.svc.cluster.local:3100/'
      querierUrl: 'http://loki-microservices-query-frontend.netobserv.svc.cluster.local:3100/'