阿里云上部署开源PaaS平台Cloud Foundry实战

2021-07-09 14:42:11

Cloud Foundry介绍

Cloud Foundry是业界第一个开源的PaaS，号称工业界唯一的云应用平台。
本文重点讲述Cloud Foundry在阿里云上的部署方案，对于Cloud Foundry介绍的部分简单略过。

Cloud Foundry解决了什么问题？

关键词：PaaS、开源
Cloud Foundry是一种平台即服务（PaaS），兼容多种基础设施云，提供多种开发框架和应用服务。目前开源支持部署Cloud Foundry的基础设施云包括AWS、Azure、OpenStack等等，CF在开发框架上支持Java、.NET、Ruby等等，有很强的灵活性。
IaaS帮助开发者和客户解决了基础设施的问题，Cloud Foundry作为PaaS，在云上让开发者的视角更高，开发者只需要关注应用和数据。
阿里云上部署开源PaaS平台Cloud Foundry实战

Cloud Foundry组件（v2版本）

Cloud Foundry是由相对独立的多个模块构成的分布式系统，每个模块单独存在和运行，各模块之间通过消息机制进行通信。
阿里云上部署开源PaaS平台Cloud Foundry实战

Cloud Foundry目前最新的架构为v3版本，各个组件的功能，官网有详细的介绍，这边不再赘述。
参考：http://docs.cloudfoundry.org/concepts/architecture/

阿里云上部署Cloud Foundry实战

环境说明：

（本文使用软件版本和环境的说明，涉及到的概念下文会详细解释）
阿里云region：华东1
ruby 版本：2.3.0
bosh-init 版本：0.0.96
Bosh CLI 版本：1.3262.4.0
CF CLI 版本：6.21.0

部署流程

在阿里云上部署Cloud Foundry，部署流程分为三大部分，分别为：

  1. 开通阿里云环境资源：包括准备阿里云账号，开通相关云产品资源等等
  2. 部署Bosh：安装bosh-init，使用bosh-init部署Bosh
  3. 部署Cloud Foundry：使用Bosh部署Cloud Foundry

本文后续内容会根据三步部署流程详细讲述。

一、开通阿里云环境资源

Cloud Foundry需要部署在VPC环境下，需要开通VPC资源，并创建虚拟交换机，规划好部署Cloud Foundry的内网网段。
同时，也需要创建一对Access Key ID和Access Key Secret，也可以使用现有的Access Key ID和Access Key Secret对。

创建专有网络VPC

网段没有限制，笔者选择了10.0.0./8网段
阿里云上部署开源PaaS平台Cloud Foundry实战

创建虚拟交换机

可用区和网段没有限制，方便起见，笔者选择了华东1可用区E，网段为10.0.0.0/25
阿里云上部署开源PaaS平台Cloud Foundry实战

创建Access Key ID和Access Key Secret

阿里云上部署开源PaaS平台Cloud Foundry实战

二、部署Bosh

Bosh介绍

Bosh是一个统一了平台即服务软件（如Cloud Foundry）的发布、部署和生命周期管理的自动化配置部署工具。简单的说，Bosh的作用之一就是部署Cloud Foundry，部署Cloud Foundry之前，首先要部署Bosh。
在讲怎么在阿里云上部署Bosh之前，首先需要了解Bosh中的几个基本概念。
Bosh官网参考：https://bosh.io/docs

CPI

CPI全称Cloud Platform Interface，是Bosh对外开放的一组标准化接口，在IaaS上部署Bosh，需要实现这组接口，Bosh用CPI对IaaS的资源进行管理，包括创建虚拟机，释放虚拟机，等等......
在阿里云上部署Bosh，需要用阿里云OpenAPI实现CPI。
CPI参考：https://bosh.io/docs/cpi-api-v1.html
CPI API v1完整接口列表：
阿里云上部署开源PaaS平台Cloud Foundry实战

bosh-init

bosh-init是Bosh官网提供的一个开源工具，它的作用就是用来部署Bosh。
bosh-init参考：https://bosh.io/docs/using-bosh-init.html

stemcell

stemcell是虚拟机镜像，预装了部署过程中需要的组件（其中最重要的组件是Bosh Agent），官方的描述是：A stemcell is a versioned Operating System image wrapped with IaaS specific packaging. Bosh用CPI创建的虚拟机，用stemcell镜像启动。
stemcell参考：https://bosh.io/docs/stemcell.html

release

release是Bosh中一个安装部署包的概念，包含了所有安装分布式系统需要的源代码、配置文件、脚本文件等等，官网描述为：A release is a versioned collection of configuration properties, configuration templates, start up scripts, source code, binary artifacts, and anything else required to build and deploy software in a reproducible way.
例如，安装Bosh，我需要一个Bosh-release；用Bosh安装Cloud Foundry，我需要一个cf-release。
所有release都有版本迭代，都可以在Bosh官网找到。
release参考：https://bosh.io/docs/release.html

Deployment

一个Deployment是一组VM的集合，由指定的stemcell镜像启动，用于部署对应release的系统。官方描述为：A deployment is a collection of VMs, built from a stemcell, that has been populated with specific releases and disks that keep persistent data. These resources are created based on a manifest file in the IaaS and managed by the BOSH Director, a centralized management server.
在Bosh的概念里，一个Deployment对应一个release，Deployment的具体配置，写在Deployment manifest里，在部署过程中提供。
Deployment参考：https://bosh.io/docs/deployment.html

阿里云上部署Bosh

1. 创建ECS

通过阿里云控制台创建一个ECS（包年包月、按量均可），用于安装bosh-init。方便起见，下文以bosh-init指代这台ECS。

推荐配置：
规格：2核4G及以上
镜像：Ubuntu 14.04 64位
系统盘：40GB以上，高效云盘
网络类型：VPC实例，选择在上一步创建好的VPC和虚拟交换机，公网IP可以选择不分配。
阿里云上部署开源PaaS平台Cloud Foundry实战

2. 给bosh-init配置公网IP

因为bosh-init是VPC实例，需要给bosh-init绑定弹性公网IP，让bosh-init可以通过公网访问。

按需购买弹性公网IP，并给bosh-init绑定弹性公网IP
阿里云上部署开源PaaS平台Cloud Foundry实战

3. 安装bosh-init

参考文档：http://bosh.io/docs/install-bosh-init.html

  1. SSH登陆到bosh-init这台ECS
  2. 下载bosh-init，下载地址见：http://bosh.io/docs/install-bosh-init.html
  3. 执行权限
    chmod +x ~/Downloads/bosh-init-*
  4. 移动到/usr/local/bin
    sudo mv ~/Downloads/bosh-init-* /usr/local/bin/bosh-init
  5. 验证安装成功
    bosh-init -v
  6. 安装对应环境，笔者使用的是Ubuntu的机器
    sudo apt-get install -y build-essential zlibc zlib1g-dev ruby ruby-dev openssl libxslt-dev libxml2-dev libssl-dev libreadline6 libreadline6-dev libyaml-dev libsqlite3-dev sqlite3
  7. 确保已经安装Ruby 2+环境
    阿里云上部署开源PaaS平台Cloud Foundry实战

4. 下载bosh-release

bosh-release是开源的，可以从Bosh官网下载：
https://bosh.io/releases/github.com/cloudfoundry/bosh?all=1

笔者使用的是255.3版本的bosh-release
阿里云上部署开源PaaS平台Cloud Foundry实战

5. 下载cpi-release

CPI近期会开源，请关注：
https://github.com/alibaba/opstools

6. 配置manifest

根据以下模板，新建一个manifest文件bosh.yml，在模板中填充阿里云资源相关的内容。

---
name: bosh

releases:
- name: bosh
  url: file:///root/downloads/bosh-255.3.tgz
- name: bosh-aliyun-cpi
  url: file:///root/downloads/bosh-aliyun-cpi.tgz

resource_pools:
- name: vms
  network: private
  cloud_properties:
    instance_type: ecs.n4.large   # <--- 实例规格

networks:
- name: private
  type: manual
  subnets:
  - range: 10.0.0.0/8
    gateway: 10.0.0.1
    cloud_properties: {
      SecurityGroupId: SECURITY_GROUP_ID,    # <--- 安全组ID
      VSwitchId: VSWITCH-ID    # <--- 虚拟交换机ID
    }
- name: public
  type: vip

jobs:
- name: bosh
  instances: 1

  templates:
  - {name: nats, release: bosh}
  - {name: redis, release: bosh}
  - {name: postgres, release: bosh}
  - {name: blobstore, release: bosh}
  - {name: director, release: bosh}
  - {name: health_monitor, release: bosh}
  - {name: registry, release: bosh}
  - {name: aliyun_cpi, release: bosh-aliyun-cpi}

  resource_pool: vms

  networks:
  - name: private
    static_ips: [10.0.0.2]    # <--- ECS内网IP
    default: [dns, gateway]
  - name: public
    static_ips: [STATIC_IP]   # <--- 弹性公网IP

  properties:
    nats: &nats
      address: 127.0.0.1
      user: nats
      password: nats-password

    redis:
      listen_address: 127.0.0.1
      address: 127.0.0.1
      password: redis-password

    postgres: &db
      listen_address: 127.0.0.1
      host: 127.0.0.1
      user: postgres
      password: postgres-password
      database: bosh
      adapter: postgres

    registry: &registry
      address: 127.0.0.1
      host: 127.0.0.1
      db: *db
      http: {user: admin, password: admin, port: 25777}
      username: admin
      password: admin
      port: 25777

    blobstore: &blobstore
      address: 127.0.0.1
      port: 25250
      provider: dav
      director: {user: director, password: director-password}
      agent: {user: agent, password: agent-password}


    director:
      address: 127.0.0.1
      name: my-bosh
      db: *db
      cpi_job: aliyun_cpi
      max_threads: 10
      user_management:
        provider: local
        local:
          users:
          - {name: admin, password: admin}
          - {name: hm, password: hm-password}

    hm:
      director_account: {user: hm, password: hm-password}
      resurrector_enabled: true

    aliyun: &aliyun
      access_key_id: ACCESS_KEY_ID # <--- 阿里云 Access Key ID
      access_key: ACCESS_KEY # <--- 阿里云 Access Key Secret
      default_key_name: bosh
      default_security_groups: [bosh]
      region_id: cn-hangzhou  # <--- 阿里云 Region

    ntp: &ntp [0.pool.ntp.org, 1.pool.ntp.org]

cloud_provider:
  template: {name: aliyun_cpi, release: bosh-aliyun-cpi}

  mbus: "https://mbus:mbus-password@10.0.0.2:6868" # <--- ECS内网IP

  properties:
    aliyun: *aliyun
    agent:
      mbus: "nats://nats:nats-password@10.0.0.2:4222"  # <--- ECS内网IP
      blobstore:
        provider: "dav"
        options:
          endpoint: "http://10.0.0.2:25250"  # <--- ECS内网IP
          user: "agent"
          password: "agent-password"
    blobstore: {provider: local, path: /var/vcap/micro_bosh/data/cache}
    ntp: *ntp
    nats: *nats
    registry: *registry
    blobstore: *blobstore

7. 部署Bosh

执行部署命令：
bosh-init deploy bosh.yml

一个部署过程的示例：
阿里云上部署开源PaaS平台Cloud Foundry实战

8. 验证Bosh

我们用Bosh CLI验证Bosh是否部署成功。
Bosh CLI是Bosh官方提供的，用于和Bosh交互的命令行工具。在部署完成Bosh之后，用Bosh CLI和Bosh交互，执行相关命令，进行下一步Cloud Foundry的部署。

安装Bosh CLI

Bosh CLI可以安装在任意一台ECS上，或者本地主机上。
安装Bosh CLI：https://bosh.io/docs/bosh-cli.html

验证Bosh

执行Bosh CLI命令：
bosh target 10.0.0.2
笔者的Bosh CLI安装在同一个安全组的ECS上，因此可以和部署Bosh的ECS进行内网通讯，直接通过内网IP，target到Bosh。如果需要通过公网通讯，需要给部署Bosh的ECS绑定弹性公网IP，或者使用NAT网关产品，保证网络能通。

如图显示，成功连接到目标Bosh，验证Bosh成功。
bosh releases、bosh stemcells这两条命令，使用Bosh部署Cloud Foundry的时候会用到，我们下节细讲。
阿里云上部署开源PaaS平台Cloud Foundry实战

三、部署Cloud Foundry

基本概念

首先我们也需要了解用Bosh部署Cloud Foundry过程中的几个基本概念。

Bosh CLI

上一节讲到，我们需要使用已经部署成功的Bosh来部署Cloud Foundry，通过Bosh CLI和Bosh进行交互，执行相关部署命令。因此，使用Bosh部署Cloud Foundry之前，首先需要了解Bosh CLI命令的使用。

Bosh CLI几个基本命令：

1.连接到指定Bosh
bosh target [DIRECTOR_URL]

2.列出当前release仓库中所有的release
bosh releases

3.上传release到Bosh的release仓库，只有仓库里的release可以用于部署
bosh upload release [RELEASE_FILE]

4.列出当前stemcell仓库里的所有stemcell
bosh stemcells

5.上传stemcell到Bosh的stemcell仓库，只有仓库里的stemcell可以用于部署
bosh upload stemcell STEMCELL_PATH
bosh upload stemcell STEMCELL_URL

6.列出当前所有的Deployment
bosh deployments

7.列出当前Deployment的信息
bosh deployment

8.切换到指定manifest对应的Deployment
bosh deployment [MANIFEST_PATH]

9.执行部署当前Deployment
bosh deploy

Bosh CLI官网参考：http://bosh.io/docs/sysadmin-commands.html

CPI

在上一步部署Bosh中，已经部署好的Bosh中已经包含了CPI组件，因此在部署Cloud Foundry中不再需要CPI-release

stemcell

和上一步部署Bosh中类似。

cf-release

Cloud Foundry官网提供的压缩包，包含Cloud Foundry所有组件的源码。

Deployment manifest

Deployment的配置文件，主要描述了用哪个stemcell，用哪个cf-release，需要部署哪些CF组件，需要多少VM，VM规格信息，VM的IP信息，具体哪个VM部署哪个CF组件，等等......

部署Cloud Foundry流程

1. 下载cf-release

cf-release是开源的，可以从Bosh官网下载：
http://bosh.io/releases/github.com/cloudfoundry/cf-release?all=1

笔者部署用的是215版本的cf-release
阿里云上部署开源PaaS平台Cloud Foundry实战

2. 登陆到安装有Bosh CLI的机器

我们需要用Bosh CLI和Bosh进行交互，执行部署Cloud Foundry命令

3. 用Bosh CLI连接到部署好的Bosh

bosh target 10.0.0.2

4. 上传cf-release

上传下载完成的cf-release，执行以下Bosh CLI命令：
bosh upload release <release_file>
阿里云上部署开源PaaS平台Cloud Foundry实战

5. 配置manifest

根据以下模板，新建一个manifest文件cf.yml，在模板中填充阿里云资源相关的内容。

---
name: ali-cf
director_uuid: BOSH_DIRECTOR_UUID  # <--- Bosh Director UUID

releases:
- {name: cf, version: 215}

networks:
- name: private
  type: manual
  subnets:
  - range: 10.0.0.0/8
    gateway: 10.0.0.1
    #dns: [10.0.0.2]
    reserved: ["10.0.0.2"]
    static: ["10.0.0.3 - 10.0.0.100"]
    cloud_properties:
      SecurityGroupId: SECURITY_GROUP_ID  # <--- 安全组ID
      VSwitchId: VSWITCH_ID  # <--- 虚拟交换机ID

resource_pools:
- name: small_ecs
  network: private
  cloud_properties:
    instance_type: ecs.n1.tiny  # <--- 实例规格

compilation:
  workers: 1
  network: private
  reuse_compilation_vms: true
  cloud_properties:
    instance_type: ecs.n1.medium  # <--- 实例规格

update:
  canaries: 1
  max_in_flight: 1
  serial: false
  canary_watch_time: 30000-600000
  update_watch_time: 5000-600000

jobs:
- name: nats
  instances: 1
  resource_pool: small_ecs
  templates:
  - {name: nats, release: cf}
  networks:
  - name: private
    static_ips: [10.0.0.20]  # <--- ECS内网IP

- name: nfs
  instances: 1
  persistent_disk: 50
  resource_pool: small_ecs
  templates:
  - {name: debian_nfs_server, release: cf}
  networks:
  - name: private
    static_ips: [10.0.0.21]  # <--- ECS内网IP

- name: postgres
  instances: 1
  persistent_disk: 50
  resource_pool: small_ecs
  templates:
  - {name: postgres, release: cf}
  networks:
  - name: private
    static_ips: [10.0.0.22]  # <--- ECS内网IP
  update:
    serial: true

- name: cloud_controller
  instances: 2
  resource_pool: small_ecs
  templates:
  - {name: cloud_controller_ng, release: cf}
  - {name: cloud_controller_worker, release: cf}
  - {name: cloud_controller_clock, release: cf}
  - {name: nfs_mounter, release: cf}
  networks:
  - name: private
    static_ips: [10.0.0.23, 10.0.0.33]  # <--- ECS内网IP
  propertis:
    nfs_server:
      address: 10.0.0.21
      allow_from_entries: [10.0.0.0/24]

- name: hm9000
  instances: 1
  resource_pool: small_ecs
  templates:
  #  - {name: consul_agent, release: cf}
  - {name: hm9000, release: cf}
  #  - {name: metron_agent, release: cf}
  - {name: route_registrar, release: cf}
  networks:
  - name: private
    static_ips: [10.0.0.24]  # <--- ECS内网IP

- name: doppler_z1
  instances: 1
  resource_pool: small_ecs
  templates:
  - {name: doppler, release: cf}
  networks:
  - name: private
  properties:
    doppler: {zone: z1}
    doppler_endpoint:
      shared_secret: PASSWORD

- name: loggregator_trafficcontroller_z1
  instances: 1
  resource_pool: small_ecs
  templates:
  - {name: loggregator_trafficcontroller, release: cf}
  - {name: metron_agent, release: cf}
  - {name: route_registrar, release: cf}
  networks:
  - name: private
  properties:
    traffic_controller: {zone: z1}
    route_registrar:
      routes:
      - name: doppler
        registration_interval: 20s
        port: 8081
        uris:
        - "doppler.REPLACE_WITH_SYSTEM_DOMAIN"
      - name: loggregator
        registration_interval: 20s
        port: 8080
        uris:
          - "loggregator.REPLACE_WITH_SYSTEM_DOMAIN"

- name: uaa
  instances: 1
  resource_pool: small_ecs
  templates:
  - {name: uaa, release: cf}
  networks:
  - name: private
    static_ips: [10.0.0.25]  # <--- ECS内网IP
  properties:
    login:
      catalina_opts: -Xmx768m -XX:MaxPermSize=256m
    uaa:
      admin:
        client_secret: PASSWORD
      batch:
        password: PASSWORD
        username: batch_user
      cc:
        client_secret: PASSWORD
      scim:
        userids_enabled: false
        users:
        - cps@aliyun.com|Cps123456|scim.write,scim.read,openid,cloud_controller.admin
    uaadb:
      address: 10.0.0.22  # <--- postgres组件内网IP
      databases:
      - {name: uaadb, tag: uaa}
      db_scheme: postgresql
      port: 5524
      roles:
      - {name: uaaadmin, password: uaa-password, tag: admin}

- name: router
  instances: 1
  resource_pool: small_ecs
  templates:
  - {name: gorouter, release: cf}
  networks:
  - name: private
    static_ips: [10.0.0.27]  # <--- ECS内网IP
  properties:
    dropsonde: {enabled: true}

- name: dea_ng
  instances: 1
  resource_pool: small_ecs
  templates:
  - {name: dea_next, release: cf}
  networks:
  - name: private
    static_ips: [10.0.0.26]  # <--- ECS内网IP

properties:
  networks: {apps: private}
  app_domains: [DOMAIN]  # <--- domain
  domain: DOMAIN  # <--- domain
  system_domain: DOMAIN  # <--- domain
  system_domain_organization: default_organization

  cc:
    allow_app_ssh_access: false
    bulk_api_password: PASSWORD
    db_encryption_key: PASSWORD
    default_running_security_groups: [public_networks, dns]
    default_staging_security_groups: [public_networks, dns]
    install_buildpacks:
    - {name: java_buildpack, package: buildpack_java}
    - {name: ruby_buildpack, package: buildpack_ruby}
    - {name: nodejs_buildpack, package: buildpack_nodejs}
    - {name: go_buildpack, package: buildpack_go}
    - {name: python_buildpack, package: buildpack_python}
    - {name: php_buildpack, package: buildpack_php}
    - {name: staticfile_buildpack, package: buildpack_staticfile}
    - {name: binary_buildpack, package: buildpack_binary}
    internal_api_password: PASSWORD
    quota_definitions:
      default:
        memory_limit: 102400
        non_basic_services_allowed: true
        total_routes: 1000
        total_services: -1
    security_group_definitions:
    - name: private
      rules: []
    srv_api_uri: http://api.DOMAIN  # <--- domain
    staging_upload_password: PASSWORD
    staging_upload_user: staging_upload_user

  ccdb:
    address: 10.0.0.22  # <--- postgres组件内网IP
    databases:
    - {name: ccdb, tag: cc}
    db_scheme: postgres
    port: 5524
    roles:
    - {name: ccadmin, password: cc-password, tag: admin}

  databases:
    databases:
    - {name: ccdb, tag: cc, citext: true}
    - {name: uaadb, tag: uaa, citext: true}
    port: 5524
    roles:
    - {name: ccadmin, password: cc-password, tag: admin}
    - {name: uaaadmin, password: uaa-password, tag: admin}
  dea_next:
    advertise_interval_in_seconds: 5
    heartbeat_interval_in_seconds: 10
  
  etcd:
    machines: [10.0.0.24]  # <--- hm9000组件内网IP

  hm9000:
    url: http://hm9000.DOMAIN  # <--- domain
    
  nats:
    machines: [10.0.0.20]  # <--- nats组件内网IP
    password: nats-password
    port: 4222
    user: nats

  nfs_server:
    no_root_squash: true
    address: 10.0.0.21  # <--- nfs组件内网IP
    allow_from_entries: [10.0.0.0/24]
  
  uaa:
    no_ssl: true
    clients:
      gorouter:
        authorities: clients.read,clients.write,clients.admin,route.admin,route.advertise
        authorized-grant-types: client_credentials,refresh_token
        scope: openid,cloud_controller_service_permissions.read
        secret: PASSWORD
      cloud_controller_username_lookup:
        authorities: scim.userids
        authorized-grant-types: client_credentials
        secret: PASSWORD
      login:
        authorities: oauth.login,scim.write,clients.read,notifications.write,critical_notifications.write,emails.write,scim.useridsassword.write
        authorized-grant-types: authorization_code,client_credentials,refresh_token
        override: true
        redirect-uri: http://login.DOMAIN:8080/auth/cloudfoundry/callback  # <--- domain
        scope: openid,oauth.approvals,cloud_controller.read,cloud_controller.write
        secret: PASSWORD

    url: http://uaa.DOMAIN  # <--- domain

6. 新建cf.yml对应的Deployment

bosh deployment cf.yml

7. 执行部署命令

bosh deploy

8. 验证Cloud Foundry

部署完成之后，执行以下命令，查看Cloud Foundry部署详情：
bosh vms
阿里云上部署开源PaaS平台Cloud Foundry实战

至此，Cloud Foundry在阿里云上部署成功。

上一篇：编译squid之文件描述符


下一篇：实践：基于深度学习的卫星图像超分辨率处理技术（一）