本文主要介绍下单机环境下zookeeper的安装，分为两个步骤：java环境搭建和zookeeper环境搭建，方便起见，所有源码放了一份在百度云上。
1. 搭建java环境
从oracle官网下载jdk.x.tar.gz文件并解压，我选择的版本是jdk1.8.0_102。
配置环境变量，修改.bashrc
JAVA_HOME=刚才的解压路径
PATH=$JAVA_HOME/bin:$JAVA_HOME/jre/bin:$PATH
CLASSPATH=.:$JAVA_HOME/lib/dt.jar:$JAVA_HOME/lib/tools.jar:$JAVA_HOME/lib:$JAVA_HOME/jre/lib:$JAVA_HOME/jre/lib/rt.jar:$CLASSPATH
source ~/.bashrc生效
验证是否安装成功：
public class test {
public static void main(String args[]) {
System.out.println("A new jdk test!");
}
}
将上述代码存为test.java，运行javac test.java生成test.class文件，执行java test如果能够输出A new jdk test!则表明java安装成功。
2. 搭建zookeeper环境
从官网镜像
下载zookeeper-x.tar.gz文件并解压，为了描述方便，假设解压后的目录路径为/home/yingshin/zookeeper，我这里选择的是zookeeper-3.4.8。
2.1 创建数据及日志目录
mkdir -p /home/yingshin/zookeeper/data
mkdir -p /home/yingshin/zookeeper/log
2.2 配置zookeeper
cp /home/yingshin/zookeeper/conf/zoo_sample.cfg /home/yingshin/zookeeper/conf/zoo.cfg
vim /home/yingshin/zookeeper/conf/zoo.cfg
修改如下配置:
tickTime=2000 #zookeeper 服务器之间或者服务器与客户端之间心跳时间，每隔该时间发送一个心跳，单位为ms
initLimit=10 #投票选举新leader的初始化时间
syncLimit=5 #leader与follower心跳检测最大容忍时间，响应超过tickTime*syncLimit，认为leader 丢失该follower
clientPort=2181 #客户端连接服务端的端口，zookeepr会监听这个端口，接收客户端的访问请求
dataDir=/home/yingshin/zookeeper/data #数据目录
dataLogDir=/home/yingshin/zookeeper/log #日志目录
2.3 配置日志
vim /home/yingshin/zookeeper/bin/zkEnv.sh 修改为：
if [ "x${ZOO_LOG_DIR}" = "x" ]
then
ZOO_LOG_DIR="/home/yingshin/zookeeper/log"
fi
if [ "x${ZOO_LOG4J_PROP}" = "x" ]
then
ZOO_LOG4J_PROP="INFO,ROLLINGFILE"
fi
2.4 配置环境变量，修改.bashrc
ZOOKEEPER_HOME=/home/yingshin/zookeeper
PATH=$ZOOKEEPER_HOME/bin:$PATH
source .bashrc生效
2.5 启停zookeeper服务
zkServer.sh负责服务的启停，启停时可以指定启动的cfg配置，默认为zoo.cfg
#指定cfg启动zookeeper
$ sh zkServer.sh start ../conf/zoo.cfg
ZooKeeper JMX enabled by default
Using config: ../conf/zoo.cfg
Starting zookeeper ... STARTED
#指定cfg查看zookeeper服务状态
$ sh zkServer.sh status ../conf/zoo.cfg
ZooKeeper JMX enabled by default
Using config: ../conf/zoo.cfg
Mode: standalone
#指定cfg停止zookeeper服务
$ sh zkServer.sh stop./conf/zoo.cfg
ZooKeeper JMX enabled by default
Using config: ../conf/zoo.cfg
Stopping zookeeper ... STOPPED
启动时如果看到以上输出，则表示zookeeper服务已经正常启动，可以通过netstat -anp | grep 2181验证zookeeper是否已经监听配置端口。
3. zk-smoketest
smoketest项目用于检测服务集群的可用性、延迟等。
$ python zk-smoketest.py --servers "127.0.0.1:2181"
Connecting to 127.0.0.1:2181
Connected in 7 ms, handle is 0
Connecting to 127.0.0.1:2181
Connected in 7 ms, handle is 1
Connecting to 127.0.0.1:2181
Connected in 15 ms, handle is 0
Smoke test successful
$ python zk-latencies.py --servers "127.0.0.1:2181" --znode_count=100
Connecting to 127.0.0.1:2181
Connected in 15 ms, handle is 0
Testing latencies on server 127.0.0.1:2181 using asynchronous calls
created 100 permanent znodes in 16 ms (0.163441 ms/op 6118.426888/sec)
set 100 znodes in 17 ms (0.175538 ms/op 5696.770163/sec)
get 100 znodes in 18 ms (0.181930 ms/op 5496.617610/sec)
deleted 100 permanent znodes in 16 ms (0.160332 ms/op 6237.068760/sec)
created 100 ephemeral znodes in 19 ms (0.190301 ms/op 5254.834749/sec)
watched 100 znodes in 16 ms (0.160351 ms/op 6236.326870/sec)
deleted 100 ephemeral znodes in 21 ms (0.211921 ms/op 4718.745359/sec)
notif 100 watches in 0 ms (included in prior)
Latency test complete
4. 参考资料：
PREVIOUS排序可视化
NEXTzookeeper介绍