ESXi服务器上利用Fuel部署Openstack错误解决

最近需要部署多节点的Openstack环境，官网上给出的手动部署文档需要输入太多命令，而且试了几次实例的网络都有各种错…
然后采用Fuel这个Openstack的自动部署工具来进行部署，首先在本地部署一遍测试下，发现在Fuel master节点部署完成之后，确实只需要点几个按钮就可以部署Openstack了。
但是毕竟要求是很多节点，本地这配置两节点还好，多了估计跑不起来，所以准备将master节点上载到ESXi服务器上，结果。。。在服务器上利用上载的master重新部署Openstack居然遇到各种错…

登录Fuel Web出错

在登录Fuel Web时报错：

unable log in, keystone service is not avaible

查看fuel master节点的keystone日志，报错：

Is the server running on host "10.20.0.2" and accepting TCP/IP connections on poer 5432?

5432端口，查了下这个端口对应的服务，貌似应该是postges服务没有启动，启动命令：

su - postgres
pg_ctl restart

再登录Fuel Web时就不报错了~~

验证网络连通性出错

错误信息

将本地fule master节点上载到服务器上后，在fuel web界面上验证网络连通性时报错：

Verification failed.
Node controller discovered DHCP server via ens192 with following parameters: IP: 10.20.0.2, MAC: 00:50:56:95:67:e0. This server will conflict with the installation.
Node controller discovered DHCP server via ens192 with following parameters: IP: 192.168.0.1, MAC: ec:6c:9f:26:47:02. This server will conflict with the installation.
Node compute-2 discovered DHCP server via ens192 with following parameters: IP: 192.168.0.1, MAC: ec:6c:9f:26:47:02. This server will conflict with the installation.
Node compute-2 discovered DHCP server via ens192 with following parameters: IP: 10.20.0.2, MAC: 00:50:56:95:67:e0. This server will conflict with the installation.
Node compute-1 discovered DHCP server via ens192 with following parameters: IP: 192.168.0.1, MAC: ec:6c:9f:26:47:02. This server will conflict with the installation.
Node compute-1 discovered DHCP server via ens192 with following parameters: IP: 10.20.0.2, MAC: 00:50:56:95:67:e0. This server will conflict with the installation.

解决方案

可以看到错误分为两块：
- 10.20.0.2与安装过程冲突
- 192.168.0.1与安装过程冲突
其中10.20.0.2是fuel master的IP，192.168.0.1是服务器网关路由的IP。
第一块错误的原因是：从本地上载master节点到服务器上，导致节点网卡的mac地址不一致。先来解决这一个错。

Fuel master节点MAC地址问题

本地的master节点四块网卡MAC地址为：

"00:0C:29:AA:51:61"
"00:0C:29:AA:51:6B"
"00:0C:29:AA:51:75"
"00:0C:29:AA:51:7F"

因此咱们需要修改服务器master四块网卡的MAC地址，修改步骤如下：

• 修改master节点对应的vmx文件：

修改前：

ethernet0.virtualDev = "e1000"
ethernet0.networkName = "network_lw_3"
ethernet0.addressType = "static"
ethernet0.present = "TRUE"
ethernet1.virtualDev = "e1000"
ethernet1.networkName = "network_lw_3"
ethernet1.addressType = "static"
ethernet1.present = "TRUE"
ethernet2.virtualDev = "e1000"
ethernet2.networkName = "network_lw_3"
ethernet2.addressType = "static"
ethernet2.present = "TRUE"
ethernet3.virtualDev = "e1000"
ethernet3.networkName = "network_lw_3"
ethernet3.addressType = "static"

uuid.bios = "42 15 69 82 80 27 f6 52-da 5c bb dd e4 b0 0d 1b"

ethernet0.address = "00:50:56:34:41:0F"
ethernet1.address = "00:0C:29:AA:51:6B"
ethernet2.address = "00:0C:29:AA:51:75"
ethernet3.address = "00:0C:29:AA:51:7F"

将网卡addressType设置为“generated”：

ethernet0.virtualDev = "e1000"
ethernet0.networkName = "network_lw_3"
ethernet0.addressType = "generated"
ethernet0.present = "TRUE"
ethernet1.virtualDev = "e1000"
ethernet1.networkName = "network_lw_3"
ethernet1.addressType = "generated"
ethernet1.present = "TRUE"
ethernet2.virtualDev = "e1000"
ethernet2.networkName = "network_lw_3"
ethernet2.addressType = "generated"
ethernet2.present = "TRUE"
ethernet3.virtualDev = "e1000"
ethernet3.networkName = "network_lw_3"
ethernet3.addressType = "generated"

VMware MAC地址的生成是根据uuid.bios生成的，并且始终保证uuid.bios的后三项即为第一块网卡的MAC地址的后三位，因此，我们需要将bios这一项需要将最后三个数字改为AA:51:61，再修改网卡MAC地址：

uuid.bios = "42 15 69 82 80 27 f6 52-da 5c bb dd e4 aa 51 61"

ethernet0.address = "00:0C:29:AA:51:61"
ethernet1.address = "00:0C:29:AA:51:6B"
ethernet2.address = "00:0C:29:AA:51:75"
ethernet3.address = "00:0C:29:AA:51:7F"

• 修改虚拟机设置中四张网卡的mac为上述配置中对应的地址：
  如第一张网卡：

• 修改vsphere，允许该端口组的MAC“地址更改”与“伪传输”

• 再验证网络，结果如下图，可以看到只有10.20.0.2的错误已经解决

网关问题

网络连通性报错的是192.168.0.1与安装过程冲突，注意我们在本地利用Fuel安装Openstack时，配置网络时，特意去掉了DHCP：

所以我们将192.168.0.1路由的DHCP功能关闭吧先….
关闭之后再重新验证，验证成功~

部署报错

成功验证网络连通性，再来部署Openstack，等待半天之后，居然又报错了！！！%>_<%

root@nat:~# cat test 
2016-11-09 07:28:31     ERR     (/Stage[main]/Osnailyfacter::Netconfig::Netconfig/Ping_host[172.18.0.1]/ensure) change from down to up failed: Timeout waiting for host '172.18.0.1' status to become 'up' after 60 seconds!
2016-11-09 07:28:31     ERR     (/Stage[main]/Osnailyfacter::Netconfig::Netconfig/Ping_host[172.18.0.1]/ensure) change from down to up failed: Timeout waiting for host '172.18.0.1' status to become 'up' after 60 seconds!
2016-11-09 07:28:31     ERR     /usr/bin/puppet:8:in `<main>'
2016-11-09 07:28:31     ERR     /usr/lib/ruby/vendor_ruby/puppet/util/command_line.rb:92:in `execute'
2016-11-09 07:28:31     ERR     /usr/lib/ruby/vendor_ruby/puppet/util/command_line.rb:146:in `run'
2016-11-09 07:28:31     ERR     /usr/bin/puppet:8:in `<main>'
2016-11-09 07:28:31     ERR     /usr/lib/ruby/vendor_ruby/puppet/util/command_line.rb:92:in `execute'
2016-11-09 07:28:31     ERR     /usr/lib/ruby/vendor_ruby/puppet/util/command_line.rb:146:in `run'
2016-11-09 07:28:31     ERR     /usr/lib/ruby/vendor_ruby/puppet/application.rb:381:in `run'
2016-11-09 07:28:31     ERR     /usr/lib/ruby/vendor_ruby/puppet/util.rb:496:in `exit_on_fail'
2016-11-09 07:28:31     ERR     /usr/lib/ruby/vendor_ruby/puppet/application.rb:381:in `block in run'
2016-11-09 07:28:31     ERR     /usr/lib/ruby/vendor_ruby/puppet/application.rb:507:in `plugin_hook'
2016-11-09 07:28:31     ERR     /usr/lib/ruby/vendor_ruby/puppet/application.rb:381:in `block (2 levels) in run'
root@nat:~# 

看错误貌似说的是172.18.0.1无响应。172.18.0.1是Fuel用到的其中一张网卡对应的网关，然而…服务器上这个网是直接配置静态IP的，根本没有网关。。。

。。

———-

更新：
最后遇到的问题是网关不通，如果是在本机上使用的VMware的话，是只需要关闭本机防火墙就可以了（因为主机本身即是这几个网段的网关）。如果是在ESXi服务器上，可能需要另外创建一台虚拟机用作NAT转换

（最后加上NAT转换测试后，虽然成功部署了Openstack，但是好像无法访问horizon……）