Skip to content

Lightweight Deployment Using the Installer

This document describes how to perform a lightweight deployment of DCE using the installer. If you want to understand what specific optimizations and reductions have been made for the lightweight version, please refer to Lightweight Deployment Optimization Validation.

Preparation

Configuration Preparation

  1. Example of manifest.yaml

    Note

    The specific component versions should match those in the offline package.

    manifest.yaml
    apiVersion: manifest.daocloud.io/v1alpha1
kind: DCEManifest
metadata:
  creationTimestamp: null
global:
  helmRepo: https://release.daocloud.io/chartrepo
  imageRepo: release.daocloud.io
infrastructures:
  istio:
    enable: true
    version: 1.16.1
  mysql:
    kpandaMode: master-slave # mgr|master-slave
    kpandaModeEnable: true
    commonMode: master-slave # mgr|master-slave
    commonModeEnable: true
    version: 8.0.29
    cpuLimit: 1
    memLimit: 2Gi
    enableAutoBackup: true
    pvcSize: 15Gi
    pvcSizeClusterPedia: 25Gi
  redis:
    version: 6.2.6-debian-10-r120
    cpuLimit: 400m
    memLimit: 500Mi
components:
    kubean:
      enable: true
      helmVersion: v0.18.4
      helmRepo: https://kubean-io.github.io/kubean-helm-chart
      variables:
    ghippo:
      enable: true
      helmVersion: 0.29.0
      variables:
    kpanda:
      enable: true
      helmVersion: 0.31.0
      variables:
    kangaroo:
      enable: true
      helmVersion: 0.20.0
      variables:
    kairship:
      enable: true
      helmVersion: 0.22.0
      variables:
    mcamel-mysql:
      enable: false
      helmVersion: 0.21.0
      variables:
    mcamel-redis:
      enable: false
      helmVersion: 0.21.0
      variables:
    mcamel-minio:
      enable: false
      helmVersion: 0.17.0
      variables:

  2. Example of clusterConfig.yaml

    clusterConfig.yaml
    apiVersion: provision.daocloud.io/v1alpha4
kind: ClusterConfig
metadata:
  creationTimestamp: null
spec:
  clusterName: my-cluster

loadBalancer:
  type: NodePort

masterNodes:
  - nodeName: "g-master1"
    ip: 10.7.10.7
    ansibleUser: "root"
    ansiblePass: "dangerous@2022"

ntpServer:
  - 0.pool.ntp.org
  - ntp1.aliyun.com
  - ntp.ntsc.ac.cn

fullPackagePath: "/home/offline"
osRepos:
  type: builtin
  isoPath: "/home/iso/Kylin-Server-V10-SP3-2403-Release-20240426-arm64.iso"
  osPackagePath: "/home/ospkgs/os-pkgs-kylin-v10sp3-v0.17.5.tar.gz"

imagesAndCharts:
  type: builtin

binaries:
  type: builtin

Offline Package Preparation

Note

The installation package must be downloaded in an online environment. Please download it in advance before arriving at the customer site.

  1. Download the offline package for the installer

    The installation package version must be v0.21.0 or higher. It is recommended to download the latest version from the official website.

    Name Version Download Updated On
    offline-v0.21.0-arm64.tar v0.21.0 ⬇ Download 2024-11-04
    offline-v0.21.0-amd64.tar v0.21.0 ⬇ Download 2024-11-04

  2. Prepare the jq tool

    Name Version Download Updated On
    jq-linux-arm64 V1.7.1 ⬇ Download 2024-11-04
    jq-linux-arm64 V1.7.1 ⬇ Download 2024-11-04

Script Preparation

Note

Please download these scripts in advance before arriving at the customer site.

  1. Istio sidecar cleanup script: clean_istio_proxy.sh

  2. Registry deployment script: mv-registry.sh

Installation Steps

Install Dependencies

Before starting the installation, make sure all prerequisite dependencies are installed.

export VERSION=v0.28.0
curl -LO https://qiniu-download-public.daocloud.io/DaoCloud_Enterprise/dce5/install_prerequisite_${VERSION}.sh
curl -LO https://qiniu-download-public.daocloud.io/DaoCloud_Enterprise/dce5/prerequisite_${VERSION}_amd64.tar.gz

export BINARY_TAR=prerequisite_${VERSION}_amd64.tar.gz
chmod +x install_prerequisite_${VERSION}.sh
./install_prerequisite_${VERSION}.sh offline full

Start Installation

For detailed installation instructions, please refer to the Installation Guide

export INSTALLER_CUSTOM_CPU_THRESHOLD=8
export INSTALLER_CUSTOM_MEM_THRESHOLD=8
export INSTALLER_COMMON_DB_SHARED=1
./dce5-installer cluster-create -d -c clusterConfig.yaml -m manifest.yml -z -j1,2,3,4,5,6,8,10,11,12

Minimal Insight Installation (Optional)

Note

Adding the Insight component requires at least 12 GB of memory.
If resources are limited, it is recommended not to install it, as this may affect the overall system stability.

To add the minimal Insight component, follow these steps:

  1. Add the following content at the end of the manifest.yaml file

    manifest.yaml
    insight:
  enable: true
  helmVersion: 0.30.0-rc2
  variables:
insight-agent:
  enable: true
  helmVersion: 0.30.0-rc2

    Note

    The specific component versions should match those in the offline package.

  2. Run steps 11 and 12 of the installer again

    ./dce5-installer cluster-create -c clusterConfig.yaml -m manifest.yml -z -j11,12

  3. Save the following as insight-server-patch.yaml

    insight-server-patch.yaml
    global:
  kafka:
    enabled: false
  tracing:
    enable: false
    aggregator:
      enabled: false
    kafkaReceiver:
      enabled: false
  applyDependenciesCrds: true
  ghippo:
    applyCR: true
  elasticAlert:
    enable: false
vector:
  enabled: false

  4. Save the following as insight-agent-patch.yaml

    insight-agent-patch.yaml
    global:
  exporters:
    trace:
      enable: false
tailing-sidecar-operator:
  enabled: false
opentelemetry-kubernetes-collector:
  enabled: false
fluent-bit:
  enabled: false

  5. Update insight-server and insight-agent

    helm upgrade -n insight-system insight offline/dce5/insight-0.30.0-rc2.tgz --version 0.30.0-rc2 --reuse-values -f insight-server-patch.yml
helm upgrade -n insight-system insight-agent offline/dce5/insight-agent-0.30.0-rc2.tgz --version 0.30.0-rc2 --reuse-values -f insight-agent-patch.yml

Post-Installation Optimization

  1. Clean up Istio sidecars

    # jq tool must be installed first
chmod +x jq-linux-amd64
mv jq-linux-amd64 /usr/local/bin/jq

# Execute sidecar cleanup script
bash clean_istio_proxy.sh

  2. Deploy registry and manage manually

    export REGISTRY_NODEPORT=30143 # Set new registry nodeport
bash mv-registry.sh clusterConfig.yaml

    Manual management steps:

    1. Global Management -> Workspaces and Folders -> Create Workspace

    2. Add authorization and set the current logged-in user as the workspace administrator

    3. Select the workspace for the image registry

    4. Repository Integration

  3. Update Multi-Cloud Management Configuration

    You need to update the seed image address of the multi-cloud management component to the new registry.

    # Suppose the current node IP is 10.5.14.160
# and the REGISTRY_NODEPORT environment variable is 30143
# Then generate a patch configuration for 10.5.14.160:30143
# * Please replace with actual IP and port values *
export REGISTRY_ADDR="10.5.14.160:30143"
cat << EOF >> kairship-patch.yml 
global:
  imageRegistry: ${REGISTRY_ADDR}/release.daocloud.io
  kairship:
    instanceConfig:
      karmadaRegistry: ${REGISTRY_ADDR}/release.daocloud.io/karmada
      kubeRegistry: ${REGISTRY_ADDR}/release.daocloud.io/karmada
EOF

# This assumes:
# The offline package extraction path is /home/offline/
# and the kairship version is 0.22.0-rc1
# * Replace with actual path and version *
export KAIRSHIP_VERSION="0.22.0-rc1"
export OFFLINE_RESOURCES_PATH="/home/offline"
helm upgrade -n kairship-system kairship \
    ${OFFLINE_RESOURCES_PATH}/dce5/kairship-${KAIRSHIP_VERSION}.tgz \
    --reuse-values -f kairship-patch.yml \
    --version ${KAIRSHIP_VERSION}

  4. Shut Down the KIND Seed Cluster

    Note

    The seed cluster should only be shut down after all installation operations have been successfully completed.

    tinder_name=$(yq '.spec.tinderKind.instanceName' clusterConfig.yaml)
[ "$tinder_name" == null ] && tinder_name=my-cluster-installer
kind delete cluster -n $tinder_name

That completes the lightweight version installation and deployment.

Comments