hue3.11.0支持s3自定义endpoint

• 背景

  HUE 3.11.0支持AWS的s3集群，但在内网环境部署时发现HUE无法访问自有的s3集群，打开s3的filebrowser后一直卡住，约6分钟后timeout。

• 分析&定位

  • 根据http://gethue.com/introducing-s3-support-in-hue/ 提供的配置指南，并无指定endpoint的配置项，只能指定region，判断是hue只会去连接Amazon的aws 集群。

  • 通过Google发现相关ISSUE：https://issues.cloudera.org/browse/HUE-5420 确认前面的分析（支持自定义endpoint的功能在3.12.0中才能支持）。

    同时由于使用的CDH版本的问题，直接升级HUE版本可能存在其他风险，决定直接在3.11.0上添加该feature。找到相关的patchset：https://github.com/cloudera/hue/commit/51e26240ac6d0aa6c86403051678bbe5d73ab8ec

    ​

• 解决方案：

  • 在aws/src/aws/conf.py中增加相关配置项的读取

    #/opt/cloudera/parcels/CDH-5.9.0-1.cdh5.9.0.p0.23/lib/hue/desktop/libs/aws/src/aws/conf.py
    def get_default_region():
        return AWS_ACCOUNTS['default'].REGION.get()
      
     +DEFAULT_CALLING_FORMAT='boto.s3.connection.S3Connection.DefaultCallingFormat'
     AWS_ACCOUNTS = UnspecifiedConfigSection(
     ...
           REGION=Config(
              key='region',
              default='us-east-1',
              type=str
     +      ),
     +      PROXY_ADDRESS=Config(
     +        help=_('Alternate address for endpoint.'),
     +        key='proxy_address',
     +        default=None,
     +        type=str
     +      ),
     +      PROXY_PORT=Config(
     +        help=_('Alternate port for endpoint.'),
     +        key='proxy_port',
     +        default=None,
     +        type=int
     +      ),
     +      CALLING_FORMAT=Config(
     +        key='calling_format',
     +        default=DEFAULT_CALLING_FORMAT,
     +        type=str
     +      ),
     +      IS_SECURE=Config(
     +        key='is_secure',
     +        default=True,
     +        type=coerce_bool
    

    ​

  • 在aws/src/aws/client.py中判断是否使用自定义endpoint然后创建对应的connection

    #/opt/cloudera/parcels/CDH-5.9.0-1.cdh5.9.0.p0.23/lib/hue/desktop/libs/aws/src/aws/client.py
      import boto
      import boto.s3
     +import boto.s3.connection
      import boto.utils
      
      from aws.conf import get_default_region, has_iam_metadata
     @@ -27,12 +28,17 @@
      
      class Client(object):
        def __init__(self, aws_access_key_id=None, aws_secret_access_key=None, aws_security_token=None, region=None,
     -               timeout=HTTP_SOCKET_TIMEOUT_S):
     +               timeout=HTTP_SOCKET_TIMEOUT_S, proxy_address=None,
     +               proxy_port=None, calling_format=None, is_secure=True):
          self._access_key_id = aws_access_key_id
          self._secret_access_key = aws_secret_access_key
          self._security_token = aws_security_token
          self._region = region.lower() if region else get_default_region()
          self._timeout = timeout
     +    self._proxy_address = proxy_address
     +    self._proxy_port = proxy_port
     +    self._calling_format = conf.DEFAULT_CALLING_FORMAT if calling_format is None else calling_format
     +    self._is_secure = is_secure
      
          boto.config.add_section('Boto')
          boto.config.set('Boto', 'http_socket_timeout', str(self._timeout))
     @@ -51,15 +57,27 @@ def from_config(cls, conf):
            aws_access_key_id=access_key_id,
            aws_secret_access_key=secret_access_key,
            aws_security_token=security_token,
     -      region=conf.REGION.get()
     +      region=conf.REGION.get(),
     +      proxy_address=conf.PROXY_ADDRESS.get(),
     +      proxy_port=conf.PROXY_PORT.get(),
     +      calling_format=conf.CALLING_FORMAT.get(),
     +      is_secure=conf.IS_SECURE.get()
          )
      
        def get_s3_connection(self):
          # First attempt to connect via specified credentials
     -    connection = boto.s3.connect_to_region(self._region,
     -                                           aws_access_key_id=self._access_key_id,
     -                                           aws_secret_access_key=self._secret_access_key)
     +    if self._proxy_address is not None and self._proxy_port is not None:
     +      # 此次有坑，需用boto.connect_s3，不要用patchset中的boto.s3.connection.S3Connection，该connection会报bad resopnse，可能是版本问题 
     +      connection = boto.connect_s3(
     +        aws_access_key_id = self._access_key_id,
     +        aws_secret_access_key = self._secret_access_key,
     +        host = self._proxy_address,
     +        port = self._proxy_port,
     +        is_secure = self._is_secure,
     +        calling_format = self._calling_format
     +        )
     +    else:
     +      connection = boto.s3.connect_to_region(self._region,
     +                                             aws_access_key_id=self._access_key_id,
     +                                             aws_secret_access_key=self._secret_access_key)
    

  • 在CDH的hue配置（hue服务范围->hue_safety_valve.ini 的 Hue 服务高级配置代码段（安全阀））中添加相关的参数proxy_address,porxy_port,is_secure,calling_format

    [aws]
    [[aws_accounts]]
    [[[default]]]
    access_key_id= your access_key_id
    secret_access_key= your secret_access_key
    allow_environment_credentials=false
    region=us-east-1
    proxy_address=10.213.44.119
    proxy_port=7480
    is_secure=false
    calling_format=boto.s3.connection.OrdinaryCallingFormat
    

  • 重启hue server 即可

    ​

• 后记

  • 如果在filebrowser中点击bucket列表进入后页面卡死，可能是因为自建的s3集群返回的文件时间戳格式不对。比如返回的时间戳格式是UTC时间 "2016-12-20T00.01.02.123Z"就会触发如下的校验，注释掉即可

    # /opt/cloudera/parcels/CDH-5.9.0-1.cdh5.9.0.p0.23/lib/hue/desktop/libs/aws/src/aws/s3/init.py 139L, 4066C
     try:
        stripped = time.strptime(datetime[:-4], '%a, %d %b %Y %H:%M:%S')
        assert datetime[-4:] == ' GMT', 'Time [%s] is not in GMT.' % datetime
      except ValueError:
        stripped = time.strptime(datetime[:-5], '%Y-%m-%dT%H:%M:%S')
       # assert datetime[-5:] == '.000Z', 'Time [%s] is not in GMT.' % datetime
      return int(calendar.timegm(stripped))
    

• Hue debug模式开关：hue服务范围->启用 Django 调试模式

  ​