构建高可靠hadoop集群之2-机栈

本文主要参考 http://hadoop.apache.org/docs/r2.8.0/hadoop-project-dist/hadoop-common/RackAwareness.html

hadoop组件是机栈敏感(译注rack,机栈，可以简单理解为节点的摆放）。

例如，HDFS块的分布会利用机栈敏感来做容错，方式是把复制放在不同的机栈。这样即便网络切换导致故障或者其它断开导致的故障，也会有一些数据是可用的。

译注：在重要的系统中，还有完全有必要考虑这个的，网络设备也会出现故障，毕竟这些网路设备的负载也很繁重的。

haoop主守护程序可以获得集群从属节点（数据节点）的机栈id，方式是激活一个完毕脚本或者是配置文件制定的java类。无论使用哪种方式，它们的输出都必须符合java的org.apache.hadoop.net.DNSToSwitchMapping接口。

这个接口要求一一对应，拓扑信息（机器摆放信息）必须形如'/myrack/myhost',其中‘/'是拓扑分隔符，’myrack'是rack的识别符，'myhost'是主机名称。假定一个rack有24个子网，那么其中一个就可以使用'/192.168.100.0/192.168.100.5'.

如果要使用java类做拓扑映射，那么类的名称必须是通过配置文件的net.topology.node.switch.mapping.impl来设定。例如,networkTopology.java,已经包含在hadoop发布程序中，管理员可以配置。

使用java类的好处是，当一个新的数据节点加入的时候，hadoop不需要调用外部进程（这样可以更高效一些）。

如果使用外部脚本，那么必须使用参数net.topology.script.file.name来配置。不同于java类，完毕拓扑脚本并没有包含在hadoop发布程序中，必须由管理员提供。当hadoop调用这些脚本的时候，会发送多个ip地址给ARGV.需要发送给脚本的ip地址个数，是

由net.topology.script.number.args控制的，默认值是100.如果net.topology.script.number.args设置为1，那么数据节点或者几点管理器每提交一个ip地址，脚本就要被调用一次。

如果net.topology.script.file.name 或者 net.topology.node.switch.mapping.impl的值没有设置，那么rack id '/default-rack'就任意ip的返回值。然而，这样的结果看起来一点也不理想，它可能会导致块同步问题（译注:大家一个rack，写入那里无所谓，所以可能性能和容错都会存在一些问题）。

译注：以上几个参数都是在core-site.xml中配置的

原文，给出了两个例子，一个是python,一个是bash

------------------------

-- python

------------------------

#!/usr/bin/python
# this script makes assumptions about the physical environment.
#  1) each rack is its own layer 3 network with a /24 subnet, which
# could be typical where each rack has its own
#     switch with uplinks to a central core router.
#
#             +-----------+
#             |core router|
#             +-----------+
#            /             \
#   +-----------+        +-----------+
#   |rack switch|        |rack switch|
#   +-----------+        +-----------+
#   | data node |        | data node |
#   +-----------+        +-----------+
#   | data node |        | data node |
#   +-----------+        +-----------+
#
# 2) topology script gets list of IP's as input, calculates network address, and prints '/network_address/ip'.

import netaddr
import sys
sys.argv.pop(0)                                                  # discard name of topology script from argv list as we just want IP addresses

netmask = '255.255.255.0'                                        # set netmask to what's being used in your environment.  The example uses a /24

for ip in sys.argv:                                              # loop over list of datanode IP's
address = '{0}/{1}'.format(ip, netmask)                      # format address string so it looks like 'ip/netmask' to make netaddr work
try:
   network_address = netaddr.IPNetwork(address).network     # calculate and print network address
   print "/{0}".format(network_address)
except:
   print "/rack-unknown"                                    # print catch-all value if unable to calculate network address

------------------------

-- bash

------------------------

#!/bin/bash
# Here's a bash example to show just how simple these scripts can be
# Assuming we have flat network with everything on a single switch, we can fake a rack topology.
# This could occur in a lab environment where we have limited nodes,like 2-8 physical machines on a unmanaged switch.
# This may also apply to multiple virtual machines running on the same physical hardware.
# The number of machines isn't important, but that we are trying to fake a network topology when there isn't one.
#
#       +----------+    +--------+
#       |jobtracker|    |datanode|
#       +----------+    +--------+
#              \        /
#  +--------+  +--------+  +--------+
#  |datanode|--| switch |--|datanode|
#  +--------+  +--------+  +--------+
#              /        \
#       +--------+    +--------+
#       |datanode|    |namenode|
#       +--------+    +--------+
#
# With this network topology, we are treating each host as a rack.  This is being done by taking the last octet
# in the datanode's IP and prepending it with the word '/rack-'.  The advantage for doing this is so HDFS
# can create its 'off-rack' block copy.
# 1) 'echo $@' will echo all ARGV values to xargs.
# 2) 'xargs' will enforce that we print a single argv value per line
# 3) 'awk' will split fields on dots and append the last field to the string '/rack-'. If awk
#    fails to split on four dots, it will still print '/rack-' last field value

echo $@ | xargs -n 1 | awk -F '.' '{print "/rack-"$NF}'

总结：

网络拓扑如何设计，无论在哪个集群中，都是需要密切关注的