Azure上Linux VM误配防火墙的恢复方法

在实际运维中，防火墙把自己挡在机器外面的情况会时有发生。如何快速的恢复对运维人员是很重要的。

本文将介绍如何用Azure Extension实现不通过ssh对VM进行操作的方法。

之前写过一遍Blog介绍如何部署Azure的CustomScriptExtension：

http://www.cnblogs.com/hengwei/p/5862200.html

在CustomScriptExtension的基础上，如果实现关闭防火墙。

1.添加防火墙规则

通过添加iptables规则关闭外部访问该VM的ssh：

iptables -A INPUT -p tcp --dport  -j DROP

所有22端口都被关闭了。

通过psping进行观察VM 22端口的情况：

psping -t 139.219.237.69:
Connecting to 139.219.237.69:: .94ms
Connecting to 139.219.237.69:: .50ms
Connecting to 139.219.237.69:: .93ms
Connecting to 139.219.237.69:: .51ms
Connecting to 139.219.237.69:: .42ms
Connecting to 139.219.237.69:: .54ms
Connecting to 139.219.237.69:: .16ms
Connecting to 139.219.237.69:: This operation returned because the timeout period expired.
Connecting to 139.219.237.69:: This operation returned because the timeout period expired.
Connecting to 139.219.237.69:: This operation returned because the timeout period expired.
Connecting to 139.219.237.69:: This operation returned because the timeout period expired.

为防止出现脚本不成功，导致再不能访问VM的情况，执行如下脚本，过5分钟自动去除防火墙：

[root@hwcentos ~]#./remove_iptables.sh &
#!/bin/bash
while true
  do
  sleep
  iptables -F
  echo `date` >> a.txt
done

2.通过CustomScriptExtension去除防火墙规则

运行PowerShell脚本：

$mycred = Get-Credential -UserName admin@xxx.partner.onmschina.cn -Message hello
Login-AzureRmAccount -EnvironmentName AzureChinaCloud -Credential $mycred
 
#定义Resource Group、VM和Location变量
$RGName = 'hwextensiontest'
$VmName = 'hwcentos'
$Location = 'China East'
 
#定义Extension相关信息
$ExtensionName = 'CustomScriptForLinux'
$Publisher = 'Microsoft.OSTCExtensions'
$version = '1.5'
 
$PublicConf = '{"commandToExecute": "iptables -F"}' 
 
#执行Set-AzureRmVMExtension命令，安装extension：
Set-AzureRmVMExtension -ResourceGroupName $RGName -VMName $VmName -Location $Location `
　　-Name $ExtensionName -Publisher $Publisher `
　　-ExtensionType $ExtensionName -TypeHandlerVersion $Version `
-Settingstring $PublicConf

RequestId IsSuccessStatusCode StatusCode ReasonPhrase
--------- ------------------- ---------- ------------
          True                OK         OK 

通过iptables -F的命令关闭所有的防火墙。

PsPing的输出结果如下：

Connecting to 139.219.237.69:: This operation returned because the timeout period expired.
Connecting to 139.219.237.69:: This operation returned because the timeout period expired.
Connecting to 139.219.237.69:: This operation returned because the timeout period expired.
Connecting to 139.219.237.69:: This operation returned because the timeout period expired.
Connecting to 139.219.237.69:: This operation returned because the timeout period expired.
Connecting to 139.219.237.69:: .12ms
Connecting to 139.219.237.69:: .16ms
Connecting to 139.219.237.69:: .64ms

这时已经可以通过ssh登录这台VM。

观察extension的日志：

[root@hwcentos 1.5.2.0]# pwd
/var/log/azure/Microsoft.OSTCExtensions.CustomScriptForLinux/1.5.2.0
[root@hwcentos 1.5.2.0]# less extension.log
..........
// :: [Microsoft.OSTCExtensions.CustomScriptForLinux-1.0]Config decoded correctly.
// :: [Microsoft.OSTCExtensions.CustomScriptForLinux-1.0]Will try to download files, number of retries = , wait SECONDS between retrievals = 20s
// :: [Microsoft.OSTCExtensions.CustomScriptForLinux-1.0]Command to execute:iptables -F
// :: [Microsoft.OSTCExtensions.CustomScriptForLinux-1.0]fileUris value provided is empty or invalid. Continue with executing command...
// :: [Microsoft.OSTCExtensions.CustomScriptForLinux-1.0]Succeeded to download files, retry count =
// :: [Microsoft.OSTCExtensions.CustomScriptForLinux-1.0]Internal DNS is ready, retry count =
// :: [Microsoft.OSTCExtensions.CustomScriptForLinux-1.0]Command to execute:iptables -F
// :: [Microsoft.OSTCExtensions.CustomScriptForLinux-1.0]Command is finished.
// :: ---stdout---
// ::
// :: ---errout---
// ::
// ::
// :: [Microsoft.OSTCExtensions.CustomScriptForLinux-1.0]Daemon,success,,Command is finished.
// :: ---stdout---
// ::
// :: ---errout---
// ::

查看防火墙状态：

[root@hwcentos ~]# iptables -L
Chain INPUT (policy ACCEPT)
target     prot opt source               destination         

Chain FORWARD (policy ACCEPT)
target     prot opt source               destination         

Chain OUTPUT (policy ACCEPT)
target     prot opt source               destination  

防火墙已经关闭。

3.延展

有时客户VM没有响应的原因并不是因为防火墙，但原因并不是防火墙。此时我们可以通过收集sosreport的信息，通过复制到其他VM等方法获得这个文件后，进行分析。

当然也可以通过ping其它VM，并在另外一台VM上抓包，看这台VM是否网络存活，以进行下一步的分析和动作。

4.需要注意的问题

当多次执行CustomScript时，这个extension会判断内容是否相同，如果相同extension会认为是重复执行，脚本将不再执行。

更新：

Azure VM的Extension的通讯机制是通过HTTPS对外发起的。如下的输出：

[root@hwcentos ~]# netstat -tunp
Active Internet connections (w/o servers)
Proto Recv-Q Send-Q Local Address               Foreign Address             State       PID/Program name
tcp               10.3.0.4:                 167.220.255.53:        SYN_RECV    -
tcp               10.3.0.4:              168.63.129.16:            TIME_WAIT   -
tcp               10.3.0.4:              40.126.88.72:            TIME_WAIT   -
tcp              10.3.0.4:                 167.220.255.53:        ESTABLISHED /sshd
tcp               10.3.0.4:              40.126.88.72:            TIME_WAIT   -
tcp               10.3.0.4:              40.126.88.72:            TIME_WAIT   -
tcp               10.3.0.4:              40.126.88.72:            TIME_WAIT   -
tcp               10.3.0.4:              168.63.129.16:            TIME_WAIT   -

可以看到VM上很多到40.x.x.x的https请求。只要VM可以访问外部的HTTPS，VM Extension就可以工作。

所以，我们的INPUT方向的iptables添加：

iptables -A INPUT -m state --state ESTABLISHED,RELATED -j ACCEPT  
或
iptables -A INPUT -m state --state ESTABLISHED -j ACCEPT

可以确保出方向的流量不被防火墙挡住。

在添加如下防火墙规则后，VM extension仍然可以把iptables的规则清除：

iptables -A INPUT -s / -j DROP