集群配置文件 clusterConfig.yaml¶
此 YAML 文件包含了集群的各项配置字段,安装之前必须先配置此文件。 该文件将定义部署模式、集群节点信息等关键参数。默认位于 offline/sample/
目录。
ClusterConfig 示例¶
以下是一个 ClusterConfig 文件示例。
clusterConfig.yaml
apiVersion: provision.daocloud.io/v1alpha4
kind: ClusterConfig
metadata:
spec:
clusterName: my-cluster
# 火种节点的域名或IP,默认解析为火种节点默认网关所在网卡的IP;可手动填入IP或域名,若为域名,如果检测到无法解析,将自动建立此域名和火种节点默认IP的映射
# bootstrapNode: auto
# kind 火种集群的配置,以下为默认值
# tinderKind:
# # kind 集群的容器名称
# instanceName: my-cluster-installer
# # kind 集群挂载的主机路径
# resourcesMountPath: /home/kind
# registryPort: 443
# minioServerPort: 9000
# minioConsolePort: 9001
# chartmuseumPort: 8081
loadBalancer:
# NodePort(default), metallb, cloudLB (Cloud Controller 暂不支持)
type: metallb
istioGatewayVip: xx.xx.xx.xx/32 # 当 loadBalancer.type 是 metallb 时必填,为 DCE 提供 UI 和 OpenAPI 访问权限
insightVip: xx.xx.xx.xx/32 # 别丢弃 /32,当 loadBalancer.type 是 metallb 时必填,用作全局服务集群的 Insight 数据采集入口,子集群的 insight-agent 可以向这个 VIP 报告数据
SourceIP: auto # 默认值auto表示开启审计日志获取源IP功能,设置为false则关闭审计日志获取源IP功能
# 指定 ssh 私钥,定义后无需再定义节点的 ansibleUser、ansiblePass
# privateKeyPath: /root/.ssh/id_rsa_sample
masterNodes:
- nodeName: "g-master1" # nodeName 将覆盖 hostName,应符合 RFC1123 标准
ip: xx.xx.xx.xx
ansibleUser: "root"
ansiblePass: "dangerous"
#ansibleSSHPort: "22"
#ansibleExtraArgs: "" # "ansible_shell_executable='/bin/sh' ansible_python_interpreter='/usr/local/bin/python'" , format: "k='v' k1='v1' k2='v2' "
- nodeName: "g-master2"
ip: xx.xx.xx.xx
ansibleUser: "root"
ansiblePass: "dangerous"
#ansibleSSHPort: "22"
#ansibleExtraArgs: ""
- nodeName: "g-master3"
ip: xx.xx.xx.xx
ansibleUser: "root"
ansiblePass: "dangerous"
#ansibleSSHPort: "22"
#ansibleExtraArgs: ""
workerNodes:
- nodeName: "g-worker1"
ip: xx.xx.xx.xx
ansibleUser: "root"
ansiblePass: "dangerous"
#ansibleSSHPort: "22"
#ansibleExtraArgs: ""
nodeTaints: # 对于 7 节点模式:至少 3 个 worker 节点应打污点(仅 ES 节点),如果使用外接 ES 则不需要添加该污点
- "node.daocloud.io/es-only=true:NoSchedule"
# nodeLabels:
# daocloud.io/hostname: g-worker1
- nodeName: "g-worker2"
ip: xx.xx.xx.xx
ansibleUser: "root"
ansiblePass: "dangerous"
#ansibleSSHPort: "22"
#ansibleExtraArgs: ""
nodeTaints:
- "node.daocloud.io/es-only=true:NoSchedule"
# nodeLabels:
# daocloud.io/hostname: g-worker2
- nodeName: "g-worker3"
ip: xx.xx.xx.xx
ansibleUser: "root"
ansiblePass: "dangerous"
#ansibleSSHPort: "22"
#ansibleExtraArgs: ""
nodeTaints:
- "node.daocloud.io/es-only=true:NoSchedule"
# nodeLabels:
# daocloud.io/hostname: g-worker3
# ntpServer:
# - 0.pool.ntp.org
# - ntp1.aliyun.com
# - ntp.ntsc.ac.cn
fullPackagePath: "/root/offline" # 解压后的离线包的路径,离线模式下该字段必填
osRepos: # 操作系统软件源
# 支持 official-service(default), builtin
type: builtin
isoPath: "/root/CentOS-7-x86_64-DVD-2009.iso"
osPackagePath: "/root/os-pkgs-centos7-v0.4.4.tar.gz"
# skipValidateOSPackage: false # 跳过 ospackage 验证
# type: external
# Set the block below only if target is S3-compatible storage which need to upload files automatically(e.g. minio).
# isoPath: "/root/CentOS-7-x86_64-DVD-2009.iso"
# osPackagePath: "/root/os-pkgs-centos7-v0.4.4.tar.gz"
# externalRepoEndpoint: https://external-repo.daocloud.io
# externalRepoUsername: rootuser
# externalRepoPassword: rootpass123
# type: external
# Set the block below if target is other storage which cannot or does not need to upload automatically(e.g. nginx).
# That requires you to import the required packages(iso, os-pkgs) manually if not all the required offline resources exist.
# `centos` as CentOS, RedHat, kylin, AlmaLinux, Fedora or Openeuler
# `debian` as Debian
# `ubuntu` as Ubuntu
# externalRepoType: centos
# externalRepoURLs: ['https://extertal-repo.daocloud.io/kubean/centos/\$releasever/os/\$basearch/']
imagesAndCharts: # 镜像仓库和 Chart仓库源
# official-service(default), builtin or external
type: builtin
# type: external
# IP or domain name
# externalImageRepo: https://external-registry.daocloud.io
# Set user and password. Optional
# externalImageRepoUsername: admin
# externalImageRepoPassword: Harbor12345
# chartmuseum or harbor
# externalChartRepoType: chartmuseum
# IP or domain name
# externalChartRepo: https://external-charts.daocloud.io:8081
# Set user and password. Optional
# externalChartRepoUsername: rootuser
# externalChartRepoPassword: rootpass123
addonPackage: # 应用商店 addon 离线包,定义后会对 addon 进行离线部署
# path:
# - "/root/standard-addon-offline-package-v0.18.0-amd64.tar.gz"
# - "/root/gpu-addon-offline-package-v0.18.0-amd64.tar.gz"
binaries: # 二进制可执行文件
# official-service(default), builtin
type: builtin
# type: external
# IP or domain name
# externalRepository: https://external-binaries.daocloud.io:9000/kubean
#externalMiddlewares:
# database:
# kpanda:
# - dbDriverName: "mysql"
# # Please refer https://gorm.io/docs/connecting_to_the_database.html
# dataSourceName: "user:password@tcp(localhost:3306)/dbname"
# # readwrite(default) or readonly
# accessType: readwrite
# # The maximum number of open connections to the database.
# #maxOpenConnections: 100
# # The maximum number of connections in the idle connection pool.
# #maxIdleConnections: 10
# # The maximum amount of time a connection may be reused.
# #connectionMaxLifetimeSeconds: 3600
# # The maximum amount of time a connection may be idle.
# #connectionMaxIdleSeconds: 1800
# ghippoApiserver:
# - dbDriverName: "mysql"
# dataSourceName: "user:password@tcp(localhost:3306)/dbname"
# ghippoKeycloak:
# - dbDriverName: "mysql"
# dataSourceName: "user:password@tcp(localhost:3306)/dbname"
# ghippoAuditserver:
# - dbDriverName: "mysql"
# dataSourceName: "user:password@tcp(localhost:3306)/dbname"
# elasticsearch:
# insight:
# endpoint: "https://xx.xx.xx.xx:9200"
# insecure: false
# # basic auth
# username: "username"
# password: "password"
# kafka:
# brokers:
# - host1:9092
# - host2:9092
# # the username and password of kafka is not necessary
# username: "username"
# password: "password"
# S3Storage:
# default:
# endpoint: "xx.xx.xx.xx:9000"
# # Set if you dont want to verify the certificate.
# insecure: true
# bucket: "bucketname"
# accessKey: "YOUR-ACCESS-KEY-HERE"
# secretKey: "YOUR-SECRET-KEY-HERE"
# Examples as below. More refer to kubespray options setting documentations.
#kubeanConfig: |-
# this config will set the timezone of nodes , and it won't change timezone if this config is commented out.
# ntp_timezone: Asia/Shanghai
# # Enable recommended node sysctl settings
# node_sysctl_tuning: true
# # Extra node sysctl settings while node_sysctl_tuning is enabled
# extra_sysctl: [{ name: net.ipv4.tcp_keepalive_time, value: 700 }]
# bin_dir: /usr/local/bin
# http_proxy: ""
# https_proxy: ""
# upstream_dns_servers:
# - 8.8.8.8
# - 8.8.4.4
# docker_mount_device: /dev/sdc
# docker_storage_options: "-s overlay2 --storage-opt overlay2.size=1G"
# k8sVersion only take effect in online mode, don't set it in offline mode.
# Unless to install a non-latest k8s version with offline pkg in place.
#k8sVersion: v1.29.5
#auditConfig:
# logPath: /var/log/audit/kube-apiserver-audit.log
# logHostPath: /var/log/kubernetes/audit
# #policyFile: /etc/kubernetes/audit-policy/apiserver-audit-policy.yaml
# #logMaxAge: 30
# #logMaxBackups: 10
# #logMaxSize: 100
# #policyCustomRules: >
# # - level: None
# # users: []
# # verbs: []
# # resources: []
#network:
# cni: calico
# clusterCIDR: 10.233.64.0/18
# serviceCIDR: 10.233.0.0/18
#cri:
# criProvider: containerd
# # criVersion only take effect in online mode, don't set it in offline mode
# #criVersion: 1.7.0
# # skip provision of CRI, default false. Currently only works with docker.
# #skipProvision: false
#renewCerts:
# # there are only 2 modes of renew certs: `onetime` or `cyclical`, default value is `cyclical`.
# #mode: cyclical
# # 1. When mode is set to `cyclical`, certificate renewal will be performed on a timer in a cyclical manner.
# #mode: cyclical
# # 2. When mode is set to `onetime`, certificate renewal will be completed at once, and you can set the validity days of the certificate.
# #mode: onetime
# # valid days can be set when in `onetime` mode, default valid days is 3650.
# #oneTimeValidDays: 3650
关键字段¶
该 YAML 文件中的关键字段说明,请参阅下表。
字段 | 说明 | 默认值 |
---|---|---|
clusterName | 在 KuBean Cluster 里的全局服务集群命名 | - |
tinderKind | 火种 kind 集群配置 | - |
tinderKind.instanceName | 火种 kind 集群的容器名称 | - |
tinderKind.resourcesMountPath | kind 集群挂载的主机路径 | /home/kind |
tinderKind.registryPort | kind 集群中镜像仓库的端口 | 443 |
tinderKind.minioServerPort | kind 集群中 MinIO Server 的端口 | 9000 |
tinderKind.minioConsolePort | kind 集群中 MinIO Console 的端口 | 9001 |
tinderKind.chartmuseumPort | kind 集群中 ChartMuseum 的端口 | 8081 |
masterNodes | 全局服务集群:Master 节点列表,包括 nodeName/ip/ansibleUser/ansiblePass 几个关键字段 | - |
masterNodes.nodeName | 节点名称,将覆盖 hostName | - |
masterNodes.ip | 节点 IP | - |
masterNodes.ansibleUser | 节点账号 | - |
masterNodes.ansiblePass | 节点密码 | - |
masterNodes.ansibleSSHPort | ssh 的端口,默认为22 | 22 |
masterNodes.ansibleExtraArgs | 指定 ansible 主机清单参数 | - |
workerNodes | 全局服务集群:Worker 节点列表,包括 nodeName/ip/ansibleUser/ansiblePass 几个关键字段 | - |
privateKeyPath | kuBean 部署集群的 SSH 私钥文件路径,如果填写则不需要定义 ansibleUser、ansiblePass | - |
k8sVersion | kuBean 安装集群的 K8s 版本必须跟 KuBean 和离线包相匹配 | - |
loadBalancer.insightVip | 如果负载均衡模式是 metallb,则需要指定一个 VIP,供给全局服务集群的 insight 数据收集入口使用,子集群的 insight-agent 可上报数据到这个 VIP | - |
loadBalancer.istioGatewayVip | 如果负载均衡模式是 metallb,则需要指定一个 VIP,供给 DCE 的 UI 界面和 OpenAPI 访问入口 | - |
loadBalancer.type | 所使用的 LoadBalancer 的模式,物理环境用 metallb,POC 用 NodePort,公有云和 SDN CNI 环境用 cloudLB(暂时还未未支持 cloudLB 模式) | NodePort (default)、metallb、cloudLB (Cloud Controller) |
loadBalancer.SourceIP | 审计日志获取源IP,副作用:在节点层面无法进行负载均衡 | auto |
fullPackagePath | 解压后的离线包的路径,离线模式下该字段必填 | - |
addonPackage.path | 应用商店 addon 包本地文件系统路径 | - |
imagesAndCharts | 镜像仓库和 Chart仓库源 | - |
imagesAndCharts.externalChartRepo | 外置 Chart 仓库的 IP 或域名 | - |
imagesAndCharts.externalChartRepoPassword | 外置 Chart 仓库的密码,用于推送镜像 | - |
imagesAndCharts.externalChartRepoType | 外置 Chart 仓库的类型,取值为 chartmuseum,harbor | - |
imagesAndCharts.externalChartRepoUsername | 外置 Chart 仓库的用户名,用于推送镜像 | - |
imagesAndCharts.externalImageRepo | 指定 external 仓库的 IP 或者域名(需指定协议头) | - |
imagesAndCharts.externalImageRepoPassword | 外置镜像仓库的密码,用于推送镜像 | - |
imagesAndCharts.externalImageRepoUsername | 外置镜像仓库的用户名,用于推送镜像 | - |
imagesAndCharts.type | 镜像与 Chart 的访问模式,取值为 official-service(在线), buitin(火种内置 registry 和 chartmuseum), external(外置) | official-service |
auditConfig | k8s api-server 的审计日志配置 | 默认关闭 |
binaries | 二进制可执行文件 | - |
binaries.externalRepository | 外置二进制可执行文件仓库的访问地址,URL 形式 | - |
binaries.type | 二进制可执行文件的访问模式,取值为 official-service(在线), builtin(火种节点内置的minio) | official-service |
network.clusterCIDR | Cluster CIDR | - |
network.cni | CNI 选择,比如 Calico、Cilium | calico |
network.serviceCIDR | Service CIDR | - |
ntpServer | 可用的 NTP 服务器,供给新节点同步时间 | - |
osRepos | 操作系统软件源 | - |
osRepos.externalRepoType | 外置软件源服务的操作系统类型, 取值为 centos(所有红帽系列), debian, ubuntu | - |
osRepos.externalRepoURLs | 外置软件源的访问地址 | - |
osRepos.isoPath | 操作系统 ISO 文件的路径, type 为 builtin 时不能为空 | - |
osRepos.osPackagePath | 系统包文件的路径 ,type 为 builtin 时不能为空 | - |
osRepos.type | 操作系统软件源的访问模式,取值为 official-service(在线), builtin(火种节点内置的minio) | official-service |
kubeanConfig.ntp_timezone | 设置节点的时区,如果不配置该参数,默认按照节点中的时区 | - |
kubeanConfig.node_sysctl_tuning | 开启后默认调整全局服务集群的 Systemctl 内核参数 | false |
kubeanConfig.extra_sysctl | 设置额外的 Systemctl 内核参数 | /usr/local/bin |
externalMiddlewares | 外置中间件 | - |
externalMiddlewares.database | 外置数据库 | - |
externalMiddlewares.database.ghippoApiserver | ghippoApiserver 外置数据库的配置 | - |
externalMiddlewares.database.ghippoAuditserver | ghippoAuditserver 外置数据库的配置 | - |
externalMiddlewares.database.ghippoKeycloak | ghippoKeycloak 外置数据库的配置 | - |
externalMiddlewares.database.kpanda | kpanda 外置数据库的配置 | - |
externalMiddlewares.database.kpanda[0].accessType | kpanda 外置数据库的访问类型,取值:readwrite,readonly | readwrite |
externalMiddlewares.database.kpanda[0].driver | kpanda 外置数据库的类型,取值:mysql | mysql |
externalMiddlewares.database.kpanda[0].dataSourceName | kpanda 外置数据库的访数据源信息,用于连接数据库,可参考 Gorm 官网连接到数据库文档 | - |
externalMiddlewares.database.kpanda[0].maxOpenConnections | kpanda 外置数据库的最大连接数 | 10 |
externalMiddlewares.database.kpanda[0].maxIdleConnections | kpanda 外置数据库的最大空闲连接数 | 10 |
externalMiddlewares.database.kpanda[0].connectionMaxLifetimeSeconds | kpanda 外置数据库的最大连接生命周期 | 0 |
externalMiddlewares.database.kpanda[0].connectionMaxIdleTimeSeconds | kpanda 外置数据库的最大空闲连接生命周期 | 0 |
externalMiddleware.elasticsearch | 外置 Elasticsearch | - |
externalMiddleware.elasticsearch.insight | insight 所使用的外置 Elasticsearch 配置 | - |
externalMiddleware.elasticsearch.insight.endpoint | insight 所使用的外置 Elasticsearch 的访问地址 | - |
externalMiddleware.elasticsearch.insight.anonymous | insight 所使用的外置 Elasticsearch 的匿名访问,取值 true,false,配置为 true 时不应再填访问凭证 | false |
externalMiddleware.elasticsearch.insight.username | insight 所使用的外置 Elasticsearch 的访问用户名 | - |
externalMiddleware.elasticsearch.insight.password | insight 所使用的外置 Elasticsearch 的访问密码 | - |
externalMiddleware.kafka | 外置 kafka | - |
externalMiddleware.kafka.insight | insight 所使用的外置 kafka 配置 | - |
externalMiddleware.kafka.insight.brokers | brokers 地址 | - |
externalMiddleware.kafka.insight.username | insight 所使用的外置 kafka 的访问用户名 | 可选 |
externalMiddleware.kafka.insight.password | insight 所使用的外置 kafka 的访问密码 | 可选 |
renewCerts | 集群证书续期 | - |
renewCerts.mode | 证书续期的两种模式,支持 cyclical、onetime | - |
精简配置说明¶
离线模式下采用 builtin 方式安装
builtin 模式意味着所需的第三方软件(如 chartMusem 、Minio、Docker registry)将由安装器进行部署并提供 DCE 5.0 平台使用。
apiVersion: provision.daocloud.io/v1alpha4
kind: ClusterConfig
metadata:
creationTimestamp: null
spec:
clusterName: my-cluster
masterNodes:
- nodeName: "g-master1" # (1)!
ip: xx.xx.xx.xx
ansibleUser: "root"
ansiblePass: "dangerous"
workerNodes:
fullPackagePath: "/root/offline"
osRepos:
type: builtin # (2)!
isoPath: "/root/CentOS-7-x86_64-DVD-2009.iso"
osPackagePath: "/root/os-pkgs-centos7-v0.4.4.tar.gz"
imagesAndCharts:
type: builtin # (3)!
addonPackage:
#path:
# - "/root/standard-addon-offline-package-v0.18.0-amd64.tar.gz"
# - "/root/gpu-addon-offline-package-v0.18.0-amd64.tar.gz"
binaries:
type: builtin # (4)!
- nodeName 将覆盖 hostName,应符合 RFC1123 标准
- official-service(if omit or empty), builtin or external
- official-service(if omit or empty), builtin or external, 目前还不支持 External S3 ...... FIXME
- official-service(if omit or empty), builtin or external
离线模式下采用 external 方式安装
external 模式意味着所需的第三方软件(如 chartMusem 、Minio、Docker registry 等等)无需安装器安装,由使用者提供地址供 DCE 5.0 平台使用。
apiVersion: provision.daocloud.io/v1alpha4
kind: ClusterConfig
metadata:
creationTimestamp: null
spec:
clusterName: my-cluster
masterNodes:
- nodeName: "g-master1" # (1)!
ip: xx.xx.xx.xx
ansibleUser: "root"
ansiblePass: "dangerous"
workerNodes:
fullPackagePath: "/root/offline"
osRepos:
type: external # (2)!
isoPath: "/root/CentOS-7-x86_64-DVD-2009.iso" # (3)!
osPackagePath: "/root/os-pkgs-centos7-v0.4.4.tar.gz" # (3)!
externalRepoType: centos # (4)!
externalRepoURLs: ["https://extertal-repo.daocloud.io/centos/\$releasever/os/\$basearch/"]
imagesAndCharts:
type: external # (5)!
externalImageRepo: https://external-registry.daocloud.io # (6)!
externalImageRepoUsername: admin
externalImageRepoPassword: Harbor12345
externalChartRepoType: chartmuseum # (7)!
externalChartRepo: https://external-charts.daocloud.io:8081 # (8)!
externalChartUsername: rootuser
externalChartMuseumPassword: rootpass123
addonPackage:
path: "/root/addon-offline-full-package-v0.4.8-amd64.tar.gz"
binaries:
type: external # (2)!
externalRepository: https://external-binaries.daocloud.io:9000/kubean # (6)!
- nodeName 将覆盖 hostName,应符合 RFC1123 标准
- official-service(if omit or empty), builtin or external
- Optional only if external repo already have full required resources
centos
as CentOS, RedHat,kylin AlmaLinux or Fedora;debian
as Debian;ubuntu
as Ubuntu- official-service(if omit or empty), builtin or external. Not Support External S3 so far...... FIXME
- Optional only if external repo already have full required resources IP or domain name
- chartmuseum or harbor
- IP or domain name
在线模式采用 official-service 方式安装
official-service 模式,当使用者采用在线安装 DCE 5.0 时,DCE 5.0 平台使用的资源将从 DaoCloud 的官方仓库进行获取。
apiVersion: provision.daocloud.io/v1alpha4
kind: ClusterConfig
metadata:
creationTimestamp: null
spec:
clusterName: my-cluster
masterNodes:
- nodeName: "g-master1" # (1)!
ip: xx.xx.xx.xx
ansibleUser: "root"
ansiblePass: "dangerous"
workerNodes:
- nodeName 将覆盖 hostName,应符合 RFC1123 标准
通过命令行生成 clusterConfig 配置文件模板¶
全模式 1 节点模式¶
# 官方在线
./dce5-installer generate-config --install-mode=cluster-create --master=1 --access-type=official-service
# 官方在线简化版
./dce5-installer generate-config --master=1
# 内建离线
./dce5-installer generate-config --install-mode=cluster-create --master=1 --access-type=builtin
# 内建离线简化版
./dce5-installer generate-config --master=1 --access-type=builtin
# 扩展离线
./dce5-installer generate-config --install-mode=cluster-create --master=1 --access-type=external
# 扩展离线简化版
./dce5-installer generate-config --master=1 --access-type=external
全模式 4 节点模式¶
# 官方在线
./dce5-installer generate-config --install-mode=cluster-create --master=3 --access-type=official-service
# 官方在线简化版
./dce5-installer generate-config --master=3
# 内建离线
./dce5-installer generate-config --install-mode=cluster-create --master=3 --access-type=builtin
# 内建离线简化版
./dce5-installer generate-config --master=3 --access-type=builtin
# 扩展离线
./dce5-installer generate-config --install-mode=cluster-create --master=3 --access-type=external
# 扩展离线简化版
./dce5-installer generate-config --master=3 --access-type=external
全模式 7节点模式¶
# 官方在线
./dce5-installer generate-config --install-mode=cluster-create --master=3 --worker=3 --access-type=official-service
# 官方在线简化版
./dce5-installer generate-config --master=3 --worker=3
# 内建离线
./dce5-installer generate-config --install-mode=cluster-create --master=3 --worker=3 --access-type=builtin
# 内建离线简化版
./dce5-installer generate-config --master=3 --worker=3 --access-type=builtin
# 扩展离线
./dce5-installer generate-config --install-mode=cluster-create --master=3 --worker=3 --access-type=external
# 扩展离线简化版
./dce5-installer generate-config --master=3 --worker=3 --access-type=external
社区版¶
# 官方在线
./dce5-installer generate-config --install-mode=install-app --access-type=official-service
# 官方在线简化版
./dce5-installer generate-config --install-mode=install-app
# 内建离线
./dce5-installer generate-config --install-mode=install-app --access-type=builtin
# 扩展离线
./dce5-installer generate-config --install-mode=install-app --access-type=external