Packstack

简介

Packstack主要是由Redhat推出的用于概念验证（PoC）环境快速部署的工具。Packstack是一个命令行工具，它使用Python封装了Puppet模块，通过SSH在服务器上部署OpenStack。

Packstack支持三种运行模式：

  • 快速运行
  • 交互式运行
  • 非交互式运行

Packstack支持两种部署架构：

  • All-in-One，即所有的服务部署到一台服务器上
  • Multi-Node，即控制节点和计算机分离

因为Redhat官方有详细的使用文档，因此本文将简要介绍Packstack的快速运行以及交互式运行方式来部署All-in-One的Openstack。

部署前准备

在开始部署前，我们需要准备一台虚拟机，它的规格如下：

名称 要求
处理器 推荐2核以上
内存 推荐4G以上
磁盘 推荐20G以上
网卡 至少一块1G网卡
操作系统 CentOS7.2

在完成虚拟机的配置和启动后，在终端下输入以下指令： 

  1. $ sudo yum install -y https://www.rdoproject.org/repos/rdo-release.rpm
  2. $ sudo yum update -y
  3. $ sudo yum install -y openstack-packstack

快速运行

快速运行模式，表示用户可以对参数不做任何配置即可开始部署，用户只需要决定是单节点还是多节点的部署方式。

单节点

在packstack命令后，使用—allinonec参数在本机上部署所有服务。

  1. $ packstack --allinone

多节点

使用—install-hosts参数来运行packstack，该参数值是由一个逗号隔开的IP地址列表。

  1. $ packstack --install-hosts=CONTROLLER_ADDRESS,NODE_ADDRESSES

Packstack在部署完成后在终端上会输出以下信息：

  1.  **** Installation completed successfully ******

交互式运行

1.如果希望以交互式的方式来配置集群的部署，可以在终端下输入：

  1. # packstack

2.packstack会提示你输入一个用于保存公共密钥的路径，输入Enter，则
会使用默认的~/.ssh/id_rsa.pub：

  1. Enter the path to your ssh Public key to install on servers:

3.packstack提示输入一个默认密码，该密码将作为admin user密码，
不输入则随机生成：

  1. Enter a default password to be used. Leave blank for a randomly generated one. :

4.输入每个wsgi服务的进程数，默认等于cpu的核数：

  1. Enter the amount of service workers/threads to use for each service. Leave blank to use the default. [%{::processorcount}] :

5.确认是否要安装MariaDB数据库,默认为y：

  1. Should Packstack install MariaDB [y|n]  [y] :

6.确认是否安装Openstack组件，可以根据需要定制服务：

  1. Should Packstack install OpenStack Image Service (Glance) [y|n]  [y] :
  2. Should Packstack install OpenStack Block Storage (Cinder) [y|n]  [y] :
  3. Should Packstack install OpenStack Shared File System (Manila) [y|n]  [n] :
  4. Should Packstack install OpenStack Compute (Nova) [y|n]  [y] :
  5. Should Packstack install OpenStack Networking (Neutron) [y|n]  [y] :
  6. Should Packstack install OpenStack Dashboard (Horizon) [y|n]  [y] :
  7. Should Packstack install OpenStack Object Storage (Swift) [y|n]  [y] :
  8. Should Packstack install OpenStack Metering (Ceilometer) [y|n]  [y] :
  9. Should Packstack install OpenStack Telemetry Alarming (Aodh) [y|n]  [y] :
  10. Should Packstack install OpenStack Resource Metering (Gnocchi) [y|n]  [y] :
  11. Should Packstack install OpenStack Clustering (Sahara). If yes it'll also install Heat. [y|n]  [n] :
  12. Should Packstack install OpenStack Orchestration (Heat) [y|n]  [n] :
  13. Should Packstack install OpenStack Database (Trove) [y|n]  [n] :
  14. Should Packstack install OpenStack Bare Metal (Ironic) [y|n]  [n] :
  15. Should Packstack install OpenStack client tools [y|n]  [y] :

7.Packstack为所有服务配置NTP服务来校准系统时间，NTP的设置只对多节点有意义：

  1. Enter a comma separated list of NTP server(s). Leave plain if Packstack should not install ntpd on instances:

8.是否安装Nagios监控服务：

  1. Should Packstack install Nagios to monitor OpenStack hosts [y|n]  [y] :

9.哪些服务器在本次安装中被排除在外：

  1. Enter a comma separated list of server(s) to be excluded. Leave plain if you don't need to exclude any server.:

10.是否启用调试模式：

  1. Do you want to run OpenStack services in debug mode [y|n]  [n] :

11.指定控制器的地址:

  1. Enter the controller host  [10.211.55.8] :

12.指定计算节点的地址：

  1. Enter list of compute hosts  [10.211.55.8] :

13.指定网络节点的地址：

  1. Enter list of network hosts  [10.211.55.8] :

14.是否使用VMWare vCenter作为hypervisor和datastore的后端：

  1. Do you want to use VMware vCenter as hypervisor and datastore [y|n]  [n] :

15.指定是否使用不支持的参数，推荐使用默认设置：

  1. Enable this on your own risk. Do you want to use unsupported parameters [y|n]  [n] :

16.网卡名称是否被自动识别为子网+CIDR的格式：

  1. Should interface names be automatically recognized based on subnet CIDR [y|n]  [n] :

17.是否为每个服务器订阅Extra Packstacks for Enterprise Linux(EPEL)，建议使用默认设置：

  1. To subscribe each server to EPEL enter "y" [y|n]  [n] :

18.是否启用自定义的软件包仓库:

  1. Enter a comma separated list of URLs to any additional yum repositories to install:

19.是否启用rdo test：

  1. To enable rdo testing enter "y" [y|n]  [n] :

20.是否启用Red Hat订阅，跳过即可：

  1. To subscribe each server to Red Hat enter a username :
  2. To subscribe each server with RHN Satellite enter RHN Satellite server URL:

21.ssl证书相关的操作：

  1. Enter the filename of the SSL CAcertificate, if the CONFIG_SSL_CACERT_SELFSIGN is set to y the path will be CONFIG_SSL_CERT_DIR/certs/selfcert.crt  [/etc/pki/tls/certs/selfcert.crt] :
  2. Enter the filename of the SSL CAcertificate Key file, if the CONFIG_SSL_CACERT_SELFSIGN is set to y the path will be CONFIG_SSL_CERT_DIR/keys/selfkey.key  [/etc/pki/tls/private/selfkey.key] :
  3. Enter the path to use to store generated SSL certificates in  [~/packstackca/] :
  4. Should packstack use selfsigned CAcert. [y|n]  [y] :
  5. Enter the ssl certificates subject country.  [--] :
  6. Enter the ssl certificates subject state.  [State] :
  7. Enter the ssl certificate subject location.  [City] :
  8. Enter the ssl certificate subject organization.  [openstack] :
  9. Enter the ssl certificate subject organizational unit.  [packstack] :
  10. Enter the ssl certificaate subject common name.  [centos-7.1.shared] :
  11. Enter the ssl certificate subject admin email.  [admin@centos-7.1.shared] :

22.配置AMQP服务，默认会使用RabbitMQ作为backend，不启用身份验证和SSL：

  1. Set the AMQP service backend [rabbitmq]  [rabbitmq] :
  2. Enter the host for the AMQP service  [10.211.55.8] :
  3. Enable SSL for the AMQP service? [y|n]  [n] :
  4. Enable Authentication for the AMQP service? [y|n]  [n] :

23.配置MariaDB服务

  1. Enter the IP address of the MariaDB server  [10.211.55.8] :
  2. Enter the password for the MariaDB admin user :
  3. Confirm password :

24.配置Identity服务，包括设置数据库连接的密码，创建默认的admin,demo用户等基本操作：

  1. Enter the password for the Keystone DB access :
  2. Confirm password :
  3. Enter y if cron job for removing soft deleted DB rows should be created [y|n]  [y] :
  4. Confirm password [y|n]  [y] :
  5. Region name  [RegionOne] :
  6. Enter the email address for the Keystone admin user  [root@localhost] :
  7. Enter the username for the Keystone admin user  [admin] :
  8. Enter the password for the Keystone admin user :
  9. Confirm password :
  10. Enter the password for the Keystone demo user :
  11. Confirm password :
  12. Enter the Keystone identity backend type. [sql|ldap]  [sql] :

25.配置Image服务，包括设置数据库连接密码，glance用户密码，后端存储：

  1. Enter the password for the Glance DB access :
  2. Confirm password :
  3. Enter the password for the Glance Keystone access :
  4. Confirm password :
  5. Glance storage backend [file|swift]  [file] :

26.配置块存储服务，包括设置数据库连接密码，cinder用户和密码:

  1. Enter the password for the Cinder DB access :
  2. Confirm password :
  3. Enter y if cron job for removing soft deleted DB rows should be created [y|n]  [y] :
  4. Confirm password [y|n]  [y] :
  5. Enter the password for the Cinder Keystone access :
  6. Confirm password :
  7. Enter the Cinder backend to be configured [lvm|gluster|nfs|vmdk|netapp|solidfire]  [lvm] :
  8. Should Cinder's volumes group be created (for proof-of-concept installation)? [y|n]  [y] :
  9. Enter Cinder's volumes group usable size  [20G] :
  10. Enter y if cron job for removing soft deleted DB rows should be created [y|n]  [y] :
  11. Confirm password [y|n]  [y] :

27.配置计算服务，包括flavor,资源虚拟比，迁移，虚拟化软件等参数的设置：

  1. Should Packstack manage default Nova flavors [y|n]  [y] :
  2. Enter the CPU overcommitment ratio. Set to 1.0 to disable CPU overcommitment  [16.0] :
  3. Enter the RAM overcommitment ratio. Set to 1.0 to disable RAM overcommitment  [1.5] :
  4. Enter protocol which will be used for instance migration [tcp|ssh]  [tcp] :
  5. Enter the compute manager for nova migration  [nova.compute.manager.ComputeManager] :
  6. Enter the path to a PEM encoded certificate to be used on the https server, leave blank if one should be generated, this certificate should not require a passphrase:
  7. Enter the SSL keyfile corresponding to the certificate if one was entered:
  8. Enter the PCI passthrough array of hash in JSON style for controller eg. [{'vendor_id':'1234', 'product_id':'5678', 'name':'default'}, {...}] :
  9. Enter the PCI passthrough whitelist as array of hash in JSON style for controller eg. [{'vendor_id':'1234', 'product_id':'5678', 'name':'default'}, {...}]:
  10. The nova hypervisor that should be used. Either qemu or kvm. [qemu|kvm]  [%{::default_hypervisor}] :
  11. Confirm password [qemu|kvm]  [%{::default_hypervisor}] :

28.配置网络服务，包括从组件，接口，网络驱动等细节的设置：

  1. Enter the password for Neutron Keystone access :
  2. Confirm password :
  3. Enter the password for Neutron DB access :
  4. Confirm password :
  5. Enter the ovs bridge the Neutron L3 agent will use for external traffic, or 'provider' if using provider networks.  [br-ex] :
  6. Enter Neutron metadata agent password :
  7. Confirm password :
  8. Should Packstack install Neutron LBaaS [y|n]  [n] :
  9. Should Packstack install Neutron L3 Metering agent [y|n]  [y] :
  10. Would you like to configure neutron FWaaS? [y|n]  [n] :
  11. Would you like to configure neutron VPNaaS? [y|n]  [n] :
  12. Enter a comma separated list of network type driver entrypoints [local|flat|vlan|gre|vxlan]  [vxlan] :
  13. Enter a comma separated ordered list of network_types to allocate as tenant networks [local|vlan|gre|vxlan]  [vxlan] :
  14. Enter a comma separated ordered list of networking mechanism driver entrypoints [logger|test|linuxbridge|openvswitch|hyperv|ncs|arista|cisco_nexus|mlnx|l2population|sriovnicswitch]  [openvswitch] :
  15. Enter a comma separated  list of physical_network names with which flat networks can be created  [*] :
  16. Enter a comma separated list of physical_network names usable for VLAN:
  17. Enter a comma separated list of <tun_min>:<tun_max> tuples enumerating ranges of GRE tunnel IDs that are available for tenant network allocation:
  18. Enter a multicast group for VXLAN:
  19. Enter a comma separated list of <vni_min>:<vni_max> tuples enumerating ranges of VXLAN VNI IDs that are available for tenant network allocation  [10:100] :
  20. Enter the name of the L2 agent to be used with Neutron [linuxbridge|openvswitch]  [openvswitch] :
  21. Enter a comma separated list of supported PCI vendor devices, defined by vendor_id:product_id according to the PCI ID Repository.  [['15b3:1004', '8086:10ca']] :
  22. Set to y if the sriov agent is required [y|n]  [n] :
  23. Enter a comma separated list of interface mappings for the Neutron ML2 sriov agent:
  24. Enter a comma separated list of bridge mappings for the Neutron openvswitch plugin:
  25. Enter a comma separated list of OVS bridge:interface pairs for the Neutron openvswitch plugin:
  26. Enter a comma separated list of bridges for the Neutron OVS plugin in compute nodes. They must be included in os-neutron-ovs-bridge-mappings and os-neutron-ovs-bridge-interfaces.:
  27. Enter interface with IP to override the default tunnel local_ip:
  28. Enter comma separated list of subnets used for tunneling to make them allowed by IP filtering.:
  29. Enter VXLAN UDP port number  [4789] :

29.设置Dashboard服务，是否开启Https服务：

  1. Would you like to set up Horizon communication over https [y|n]  [n] :

30.配置对象存储服务,包括设备逻辑，zone，replicas,文件系统和块设备大小的配置：

  1. Enter the Swift Storage devices e.g. /path/to/dev:
  2. Enter the number of swift storage zones, MUST be no bigger than the number of storage devices configured  [1] :
  3. Enter the number of swift storage replicas, MUST be no bigger than the number of storage zones configured  [1] :
  4. Enter FileSystem type for storage nodes [xfs|ext4]  [ext4] :
  5. Enter the size of the storage device (eg. 2G, 2000M, 2000000K)  [2G] :

31.是否启用Tempest服务:

  1. Would you like to provision for demo usage and testing [y|n]  [y] :
  2. Would you like to configure Tempest (OpenStack test suite). Note that provisioning is only supported for all-in-one installations. [y|n]  [n] :

32.设置Floating IP网段

  1. Enter the network address for the floating IP subnet  [172.24.4.224/28] :

33.设置测试镜像的名称，源地址，格式等配置：

  1. Enter the name to be assigned to the demo image  [cirros] :
  2. Enter the location of an image to be loaded into Glance  [http://download.cirros-cloud.net/0.3.4/cirros-0.3.4-x86_64-disk.img] :
  3. Enter the format of the demo image  [qcow2] :
  4. Enter the name of a user to use when connecting to the demo image via ssh  [cirros] :
  5. Enter the name to be assigned to the uec image used for tempest  [cirros-uec] :
  6. Enter the location of a uec kernel to be loaded into Glance  [http://download.cirros-cloud.net/0.3.4/cirros-0.3.4-x86_64-kernel] :
  7. Enter the location of a uec ramdisk to be loaded into Glance  [http://download.cirros-cloud.net/0.3.4/cirros-0.3.4-x86_64-initramfs] :
  8. Enter the location of a uec disk image to be loaded into Glance  [http://download.cirros-cloud.net/0.3.4/cirros-0.3.4-x86_64-disk.img] :
  9. Would you like to configure the external ovs bridge [y|n]  [y] :

34.设置Ceilometer,Aodh,Gnocchi服务：

  1. Enter the password for Gnocchi DB access :
  2. Confirm password :
  3. Enter the password for the Gnocchi Keystone access :
  4. Confirm password :
  5. Enter the password for the Ceilometer Keystone access :
  6. Confirm password :
  7. Enter the Ceilometer service name. [ceilometer|httpd]  [httpd] :
  8. Enter the host for the MongoDB server  [10.211.55.8] :
  9. Enter the host for the Redis server  [10.211.55.8] :
  10. Enter the port of the redis server(s)  [6379] :
  11. Enter the password for the Aodh Keystone access :
  12. Confirm password :

35.设置nagios用户的密码:

  1. Enter the password for the nagiosadmin user :

36.最后一步，确认生成的配置是否符合期望，输入yes，并按回车键开始执行操作：

  1. Packstack will be installed using the following configuration:
  2. ==============================================================
  3. ssh-public-key:                /root/.ssh/id_rsa.pub
  4. default-password:
  5. service-workers:               %{::processorcount}
  6. mariadb-install:               y
  7. ......
  8. aodh-ks-passwd:                ********
  9. nagios-passwd:                 ********
  10. Proceed with the configuration listed above? (yes|no):

非交互式方式运行

使用下述命令生成一个answer file:

  1. # packstack --gen-answer-file=my_file

使用vim打开文件,每个配置项都含有详细的说明：

  1. [general]
  3. # Path to a public key to install on servers. If a usable key has not
  4. # been installed on the remote servers, the user is prompted for a
  5. # password and this key is installed so the password will not be
  6. # required again.
  7. CONFIG_SSH_KEY=/root/.ssh/id_rsa.pub
  9. # Default password to be used everywhere (overridden by passwords set
  10. # for individual services or users).
  11. CONFIG_DEFAULT_PASSWORD=
  13. # The amount of service workers/threads to use for each service.
  14. # Useful to tweak when you have memory constraints. Defaults to the
  15. # amount of cores on the system.
  16. CONFIG_SERVICE_WORKERS=%{::processorcount}
  18. # Specify 'y' to install MariaDB. ['y', 'n']
  19. CONFIG_MARIADB_INSTALL=y
  21. # Specify 'y' to install OpenStack Image Service (glance). ['y', 'n']
  22. CONFIG_GLANCE_INSTALL=y
  24. ......

例如，我们不希望配置MariaDB，只需要将CONFIG_MARIADB_INSTALL设置为n:

  1. CONFIG_MARIADB_INSTALL=n

保存并退出my_file，在终端下运行以下命令指定相应的配置文件：

  1. # packstack --answer-file=my_file

深入理解Packstack

Packstack的使用非常简单，关于如何使用的介绍就到此结束。接下来才是重点，我们要深入到Packstack的核心逻辑: Plugin，并且举例说明如何编写Plugin来完成对Packstack的功能扩展。

什么是Plugin

在前面两章对于PuppetOpenstack modules的介绍中，所有服务的部署工作实际是由每个modules完成的。
在使用Packstack的时候，我们发现Packstack支持大量的服务部署，例如:nova,glance,maridb,amqp等等。在其背后每个服务的配置项管理都是由plugin实现的，其路径是在: packstack/plugins，它看起来是这样的：

  • init.py
  • amqp_002.py
  • aodh_810.py
  • …
  • trove_850.py

每个plugin的名称都是由服务名称+下划线+三位数字编码组成，那么这些数字有什么作用？

我们来看一下packstack代码入口packstack/installer/run_setup.py是怎么加载plugins的:

在主函数入口，可以看到第一步调用了loadPlugins函数来加载插件：

  1. def main():
  2.     options = ""
  4.     try:
  5.         # Load Plugins
  6.         loadPlugins()   
  7.         initPluginsConfig()

接着，我们跳转到了loadPlugins函数的定义，可以看到其中使用了sorted函数对由plugin文件组成的列表进行排序：

  1. def loadPlugins():
  2.     """
  3.     Load All plugins from ./plugins
  4.     """
  5.     sys.path.append(basedefs.DIR_PLUGINS)
  6.     sys.path.append(basedefs.DIR_MODULES)
  8.     fileList = [f for f in os.listdir(basedefs.DIR_PLUGINS) if f[0] != "_"]
  9.     fileList = sorted(fileList, cmp=plugin_compare)  #使用plugin_compare函数作为key进行排序
  10.     for item in fileList:
  11.         # Looking for files that end with ###.py, example: a_plugin_100.py
  12.         match = re.search("^(.+\_\d\d\d)\.py$", item)
  13.         if match:
  14.             try:
  15.                 moduleToLoad = match.group(1)
  16.                 logging.debug("importing module %s, from file %s", moduleToLoad, item)
  17.                 moduleobj = __import__(moduleToLoad)
  18.                 moduleobj.__file__ = os.path.join(basedefs.DIR_PLUGINS, item)
  19.                 globals()[moduleToLoad] = moduleobj
  20.                 checkPlugin(moduleobj)
  21.                 controller.addPlugin(moduleobj)
  22.             except:
  23.                 logging.error("Failed to load plugin from file %s", item)
  24.                 logging.error(traceback.format_exc())
  25.                 raise Exception("Failed to load plugin from file %s" % item)

查看函数plugin_compare的定义，我们终于找到了关键，plugin_compare使用每个plugin文件尾缀的三位数字用于排序比较：

  1. def plugin_compare(x, y):
  2.     """
  3.     Used to sort the plugin file list
  4.     according to the number at the end of the plugin module
  5.     """
  6.     x_match = re.search(".+\_(\d\d\d)", x)
  7.     x_cmp = x_match.group(1)
  8.     y_match = re.search(".+\_(\d\d\d)", y)
  9.     y_cmp = y_match.group(1)
  10.     return int(x_cmp) - int(y_cmp)

在了解了plugin的加载顺序后，我们再看看Plugin的代码结构。实际上，每个plugin的代码结构是一致的，由两个函数组成：

  • initConfig(controller) 用于初始化Plugin的配置，主要是参数和参数组。
  • initSequences(controller) 用于定义该plugin执行的任务。

在这些plugin中，必然会有一些与众不同的plugin，比如说第一个被加载的plugin，，倒数第二个被加载的plugin，以及最后一个被加载的plugin:

  • prescript_000.py是第一个被加载的plugin，顾名思义它提供了一些全局的初始化设置，比如ssh public key，default_password，workers的进程数量，是否开启各个OpenStack服务的设置等等，同时它会在被管理的主机上执行一些预备任务：生成authorized_keys文件，安装并开启epel源和rdo源，安装puppet软件包依赖和module依赖等等。
  • puppet_950.py是一个重要的plugin，顾名思义它提供了与puppet相关的任务，例如：生成最终的manifest文件，拷贝puppet modules到指定主机，生成hieradata文件，以standalone方式运行puppet：执行puppet apply，获取puppet运行中的输出等等。
  • postscript_951.py是最后一个呗加载的plugin，它只做了一件事情，就是运行Tempest跑测试任务。

动手写一个Plugin

在了解了plugin的运行机制后，我们来动手写一个plugin，我们称之为NOOP：这是一个空Plugin，默认只输出一行信息: NOOP Plugin.

创建一个Plugin文件

在packstack/plugins目录下，我们创建一个plugin文件: noop_840.py。

设置Import和Plugin定义

  1. # -*- coding: utf-8 -*-
  3. """
  4. Installs and configures NOOP
  5. """
  7. from packstack.installer import basedefs
  8. from packstack.installer import validators
  9. from packstack.installer import processors
  10. from packstack.installer import utils
  12. from packstack.modules.common import filtered_hosts
  13. from packstack.modules.documentation import update_params_usage
  14. from packstack.modules.ospluginutils import generate_ssl_cert
  16. # ------------- NOOP Packstack Plugin Initialization --------------
  18. PLUGIN_NAME = "NOOP"
  19. PLUGIN_NAME_COLORED = utils.color_text(PLUGIN_NAME, 'blue')

每个Plugin的import可能会有所不同，但大多数都会用到packstack.installer和packstack.modules。

此外，这里有两个和plugin相关的变量：

变量 说明
PLUGIN_NAME plugin名称，全部大写字母
PLUGIN_NAME_COLORED plugin显示的颜色，默认使用blue即可。

定义Plugin的配置信息

我们定义一个initConfig函数，其中包含了两个变量:

  • params
  • group
  1. def initConfig(controller):
  2.     params = [
  3.                {"CMD_OPTION": "enable-noop",
  4.                 "USAGE": "To set up noop service set this to 'y'",
  5.                 "PROMPT": "Would you like to set up noop service",
  6.                 "OPTION_LIST": ["y", "n"],
  7.                 "VALIDATORS": [validators.validate_options],
  8.                 "DEFAULT_VALUE": "n",
  9.                 "MASK_INPUT": False,
  10.                 "LOOSE_VALIDATION": True,
  11.                 "CONF_NAME": "CONFIG_ENABLE_NOOP",
  12.                 "USE_DEFAULT": False,
  13.                 "NEED_CONFIRM": False,
  14.                 "CONDITION": False},
  15.              ]
  16.     group = {"GROUP_NAME": "NOOP",
  17.              "DESCRIPTION": "NOOP Config parameters",
  18.              "PRE_CONDITION": "CONFIG_NOOP_INSTALL",
  19.              "PRE_CONDITION_MATCH": "y",
  20.              "POST_CONDITION": False,
  21.              "POST_CONDITION_MATCH": True}
  22.     controller.addGroup(group, params)

params是NOOP plugin定义的配置项，每个配置项的数据类型是字典。这些配置项可以作为顺序执行的一部分，或者作为Puppet模板的变量。

选项 说明
CMD_OPTION 被命令行使用的选项名称
USAGE 选项的使用说明，同时作为answer file的注释
PROMPT 交互模式下给用户的提示
OPTION_LIST 可选值列表，可以设置为[]或移除该选项，表示对选项值无限制
VALIDATORS 验证器函数列表，用于检查输入是否符合要求
DEFAULT_VALUE 选项的默认值
PROCESSORS 处理器函数列表，处理器函数对用户的输入做了处理，比如processors.process_host将主机名转变为IP地址等等
MASK_INPUT 是否隐藏用户的输入，如password
LOOSE_VALIDATION 若为true，则即使验证器返回为false，仍然使用用户输入的选项值
CONF_NAME 在answer file中的配置项名称，你同时可以在controller.CONF dict中找到
USE_DEFAULT 若为true,在交互模式下，将不会要求用户输入此变量的值，而直接DEFAULT_VALUE
NEED_CONFIRM 若为true，则要求用户确认其输入（比如password)
CONDITION enable/disable该选项的条件,总是设置为False即可
DEPRECATES 弃用的CONF_NAME选项列表，通常在新版本时使用

group表示组的概念，在Packstack中，会把相关的配置项分组，这样就可以通过组的方式来管理和使用。group的数据类型是字典:

选项 说明
GROUP_NAME 组名，全局唯一
DESCRIPTION 组的描述，在命令行的帮助命令下会显示此信息
PRE_CONDITION 前提条件，可以是一个配置项的值或函数的返回值匹配预期。若为False，那么该配置组处于启用状态
PRE_CONDITION_MATCH 前提条件的预期匹配值
POST_CONDITION 配置组所有参数是正确的后置条件，若设置为False，则表示不做检查。通常设置为False
POST_CONDITION_MATCH 后置条件的预期匹配值，通常设置为True

这里最重要的是PRE_CONDITION和PRE_CONDITION_MATCH，可能有些晦涩，我们以部署Cinder服务为例，只有当PRE_CONDITION中的变量CONFIG_CINDER_INSTALL为”y”时，才会显示”Cinder”组的配置选项:

  1.     {"GROUP_NAME": "CINDER",
  2.      "DESCRIPTION": "Cinder Config parameters",
  3.      "PRE_CONDITION": "CONFIG_CINDER_INSTALL",
  4.      "PRE_CONDITION_MATCH": "y",
  5.      "POST_CONDITION": False,
  6.      "POST_CONDITION_MATCH": True
  7.     }

最后一步，把这些已定义的选项添加controller的组中：

  1. controller.addGroup(group, params)

定义函数执行顺序

前面我们说到每个plugin除了定义一组相关的选项之外，还会执行一些任务，比如：从用户给定的变量值来渲染template，从而生成该服务的Puppet manifest文件。这些任务是由一个个函数组成，函数之间有执行的先后顺序，这个顺序就是由initSeqeuence函数来决定。

我们假设NOOP服务的安装需要数据库服务MariaDB，以及在Keystone中创建endpoint等操作:

  1. def initSequences(controller):
  2.     if controller.CONF['CONFIG_NOOP_INSTALL'] != 'y':
  3.     return
  4.     steps = [{'title': 'Adding MariaDB manifest entries',
  5.              'functions': [create_mariadb_manifest]},
  6.             {'title': 'Adding NOOP manifest entries',
  7.              'functions': [create_manifest]},
  8.             {'title': 'Adding NOOP Keystone manifest entries',
  9.              'functions': [create_keystone_manifest]}]
  10. controller.addSequence('Installing NOOP service', [], [], steps)

setps是一个列表，其中每个元素的数据类型都是字典，它的格式如下：

选项 说明
title 函数的简单描述信息
functions 函数列表

最后，我们调用controller.addSequence()方法把plugin的steps添加到将要被执行的序列列表中。
通常情况下，第二个和第二个选项为空。

生成manifest文件

在讲生成manifest文件之前，我们需要花些时间来了解packstack的templates，在当前版本中packstack的templates的路径是packstack/puppet/templates，数量已经从几十个削减为3个:

  • controller
  • compute
  • network

我们以controller为例，我们选取其中的代码片段:

  1. stage { "init": before  => Stage["main"] }
  3. Exec { timeout => hiera('DEFAULT_EXEC_TIMEOUT') }
  4. Package { allow_virtual => true }
  6. class {'::packstack::prereqs':
  7.   stage => init,
  8. }
  10. include ::firewall
  12. if hiera('CONFIG_NTP_SERVERS', '') != '' {
  13.   include '::packstack::chrony'
  14. }
  16. include '::packstack::amqp'
  17. include '::packstack::mariadb'
  19. if hiera('CONFIG_MARIADB_INSTALL') == 'y' {
  20.   include 'packstack::mariadb::services'
  21. } else {
  22.   include 'packstack::mariadb::services_remote'
  23. }

我们可以发现，这实质上是一个manifest文件(.pp)，而非template文件(.erb)，所有class或define的include不再使用erb模板的方式来渲染，而是使用了简单的条件判断来做选择。

每个函数都是接受固定的两个参数:

  • config
  • messages