Zookeeper Java客户端连接慢、超时问题Ad-Hoc检查清单

Zookeeper Java客户端连接慢、超时问题Ad-Hoc检查清单

 记录臭名昭著的Zookeeper连接卡20s问题，快速定位解决Java客户端连接Zookeeper可能遇到的各种连接慢、超时问题。 

TL;DR

排查思路：

首先确认你的设备到zookeeper的连通性是OK的，可通过命令echo srvr | nc HOST 2181，检查是否可以正常打印节点信息。windows用户可以在命令行输入telnet HOST 2181连接后输入srvr然后回车。

若步骤1检查OK，但就是连接慢或者超时，则通过启动进程连接zk期间执行jstack -l <pid>获取线程dump信息进一步分析，若不想一一排查，可尝试下面方法快速试下： 

本地配置hosts文件，添加下面条目：

127.0.0.1 localhost mbpro.local

::1 localhost mbpro.local

 注意：mbpro.local要替换为hostname命令的输出

启动Java进程时配置系统属性：-Djava.net.preferIPv4Stack=true -Dzookeeper.sasl.client=false

case记录

Zookeeper Java客户端初始化打印日志记录环境信息，卡在InetAddress.getLocalHost().getCanonicalHostName()方法

"ZookeeperServiceUrlProvider-1" #1 prio=5 os_prio=0 tid=0x00000000033b0800 nid=0x432c runnable [0x000000000329e000]

   java.lang.Thread.State: RUNNABLE

 at java.net.Inet4AddressImpl.lookupAllHostAddr(Native Method)

 at java.net.InetAddress$2.lookupAllHostAddr(InetAddress.java:928)

 at java.net.InetAddress.getAddressesFromNameService(InetAddress.java:1323)

 at java.net.InetAddress.getAllByName0(InetAddress.java:1276)

 at java.net.InetAddress.getAllByName0(InetAddress.java:1253)

 at java.net.InetAddress.getHostFromNameService(InetAddress.java:634)

 at java.net.InetAddress.getCanonicalHostName(InetAddress.java:588)

 at org.apache.zookeeper.Environment.list(Environment.java:62)

 at org.apache.zookeeper.Environment.logEnv(Environment.java:98)

 at org.apache.zookeeper.ZooKeeper.<clinit>(ZooKeeper.java:97)

 at ...

解决方法：本地配置hosts文件，添加下面条目

127.0.0.1 localhost mbpro.local

::1 localhost mbpro.local

注意：mbpro.local要替换为hostname命令的输出

DNS解析响应慢导致zk客户端SendThread卡住较长时间

"ZookeeperServiceUrlProvider-1-SendThread()@9231" daemon prio=5 tid=0x1b nid=NA waiting

  java.lang.Thread.State: WAITING

   at java.lang.Object.wait(Object.java:-1)

   at java.lang.Object.wait(Object.java:502)

   at java.net.InetAddress.checkLookupTable(InetAddress.java:1393)

   at java.net.InetAddress.getAddressesFromNameService(InetAddress.java:1310)

   at java.net.InetAddress.getAllByName0(InetAddress.java:1276)

   at java.net.InetAddress.getAllByName0(InetAddress.java:1253)

   at java.net.InetAddress.getHostFromNameService(InetAddress.java:634)

   at java.net.InetAddress.getHostName(InetAddress.java:559)

   at java.net.InetAddress.getHostName(InetAddress.java:531)

   at java.net.InetSocketAddress$InetSocketAddressHolder.getHostName(InetSocketAddress.java:82)

   at java.net.InetSocketAddress$InetSocketAddressHolder.access$600(InetSocketAddress.java:56)

   at java.net.InetSocketAddress.getHostName(InetSocketAddress.java:345)

   at org.apache.zookeeper.ClientCnxn$SendThread.startConnect(ClientCnxn.java:998)

   at org.apache.zookeeper.ClientCnxn$SendThread.run(ClientCnxn.java:1060)

解决方法：

进一步排查DNS解析响应慢的问题

切换至使用IP地址方式连接

客户端尝试SASL认证方式触发DNS解析，DNS解析响应慢导致超时

"ZookeeperServiceUrlProvider-1-SendThread(HOST:2181)" #34 daemon prio=5 os_prio=0 tid=0x0000024d6f94a000 nid=0xa984 runnable [0x000000ff9cffe000]

   java.lang.Thread.State: RUNNABLE

        at java.net.Inet4AddressImpl.getHostByAddr(Native Method)

        at java.net.InetAddress$2.getHostByAddr(InetAddress.java:933)

        at java.net.InetAddress.getHostFromNameService(InetAddress.java:618)

        at java.net.InetAddress.getHostName(InetAddress.java:560)

        at java.net.InetAddress.getHostName(InetAddress.java:532)

        at java.net.InetSocketAddress$InetSocketAddressHolder.getHostName(InetSocketAddress.java:82)

        at java.net.InetSocketAddress$InetSocketAddressHolder.access$600(InetSocketAddress.java:56)

        at java.net.InetSocketAddress.getHostName(InetSocketAddress.java:345)

        at org.apache.zookeeper.SaslServerPrincipal$WrapperInetSocketAddress.getHostName(SaslServerPrincipal.java:105)

        at org.apache.zookeeper.SaslServerPrincipal.getServerPrincipal(SaslServerPrincipal.java:59)

        at org.apache.zookeeper.SaslServerPrincipal.getServerPrincipal(SaslServerPrincipal.java:41)

        at org.apache.zookeeper.ClientCnxn$SendThread.startConnect(ClientCnxn.java:1161)

        at org.apache.zookeeper.ClientCnxn$SendThread.run(ClientCnxn.java:1211)

   Locked ownable synchronizers:

        - None