You can deploy Harbor on Kubernetes via helm to make it highly available. In this way, if one of the nodes on which Harbor is running becomes unavailable, users do not experience interruptions of service.

Prerequisites

  • Kubernetes cluster 1.10+
  • Helm 2.8.0+
  • High available ingress controller (Harbor does not manage the external endpoint)
  • High available PostgreSQL database (Harbor does not handle the deployment of HA of database)
  • High available Redis (Harbor does not handle the deployment of HA of Redis)
  • PVC that can be shared across nodes or external object storage

Architecture

Most of Harbor’s components are stateless now. So we can simply increase the replica of the pods to make sure the components are distributed to multiple worker nodes, and leverage the “Service” mechanism of K8S to ensure the connectivity across pods.

As for storage layer, it is expected that the user provide high available PostgreSQL, Redis cluster for application data and PVCs or object storage for storing images and charts.

Harbor High Availability with Helm

Harbor High Availability with Helm

Download Chart

Download Harbor helm chart:

  1. helm repo add harbor https://helm.goharbor.io
  2. helm fetch harbor/harbor --untar

Configuration

Configure the followings items in values.yaml, you can also set them as parameters via --set flag during running helm install:

  • Ingress rule Configure the expose.ingress.hosts.core and expose.ingress.hosts.notary.

  • External URL Configure the externalURL.

  • External PostgreSQL Set the database.type to external and fill the information in database.external section.

    Four empty databases should be created manually for Harbor core, Clair, Notary server and Notary signer and configure them in the section. Harbor will create tables automatically when starting up.

  • External Redis Set the redis.type to external and fill the information in redis.external section.

    Harbor introduced redis Sentinel mode support in 2.1.0. You can enable this by setting sentinel_master_set and host to <host_sentinel1>:<port_sentinel1>,<host_sentinel2>:<port_sentinel2>,<host_sentinel3>:<port_sentinel3>.

    You can also refer to this guide to setup a HAProxy before the Redis to expose a single entry point.

  • Storage By default, a default StorageClass is needed in the K8S cluster to provision volumes to store images, charts and job logs.

    If you want to specify the StorageClass, set persistence.persistentVolumeClaim.registry.storageClass, persistence.persistentVolumeClaim.chartmuseum.storageClass and persistence.persistentVolumeClaim.jobservice.storageClass.

    If you use StorageClass, for both default or specified one, set persistence.persistentVolumeClaim.registry.accessMode, persistence.persistentVolumeClaim.chartmuseum.accessMode and persistence.persistentVolumeClaim.jobservice.accessMode as ReadWriteMany, and make sure that the persistent volumes must can be shared cross different nodes.

    You can also use the existing PVCs to store data, set persistence.persistentVolumeClaim.registry.existingClaim, persistence.persistentVolumeClaim.chartmuseum.existingClaim and persistence.persistentVolumeClaim.jobservice.existingClaim.

    If you have no PVCs that can be shared across nodes, you can use external object storage to store images and charts and store the job logs in database. Set the persistence.imageChartStorage.type to the value you want to use and fill the corresponding section and set jobservice.jobLogger to database.

  • Replica Set portal.replicas, core.replicas, jobservice.replicas, registry.replicas, chartmuseum.replicas, clair.replicas, notary.server.replicas and notary.signer.replicas to n(n>=2).

Installation

Install the Harbor helm chart with a release name my-release:

Helm 2:

  1. helm install --name my-release .

Helm 3:

  1. helm install my-release .