【转载】关于java dns cache (域名缓存时间)

今天刚刚碰到一个JVM缓存DNS的引起的问题，分享一下一个同事的博文：

 

------------------------------------------------------丑陋的分割线---------------------------------------------

最近手上的某java应用频繁因网络问题而出现故障，同时也抛出一个问题：JAVA本身对DNS的缓存时间是多久？

对于非公司内部产品的疑问，第一反应Google之，大致有两种说法：
第1种：默认情况下networkaddress.cache.ttl=-1，代表永久缓存（配置文件路径： JAVA_HOME/jre/lib/security/java.security），就是在应用启动之后第一次DNS 解析成功的结果会一直cache到应用停止。显然在域名对应的IP有变更的时候，如果不重启应用就会造成故障。有部分同事以前也做过相关测试，认同这种说法。

第2种：jdk1.5和1.5之前的版本默认DNS 缓存时间是永久缓存，jdk 1.6以后与security manager策略有关(jboss tomcat 等app server默认不启用，详见此文)，如果没有启用security manager ，默认DNS 缓存时间30秒。策略配置文件：JAVA_HOME/jre/lib/security/java.policy

根据上述说法，先看一下配置文件JAVA_HOME/jre/lib/security/java.security

^?View Code BASH

#
# The Java-level namelookup cache policy for successful lookups:
#
# any negative value: caching forever
# any positive value: the number of seconds to cache an address for
# zero: do not cache
#
# default value is forever (FOREVER). For security reasons, this
# caching is made forever when a security manager is set. When a security
# manager is not set, the default behavior in this implementation
# is to cache for 30 seconds.
#
# NOTE: setting this to anything other than the default value can have
#       serious security implications. Do not set it unless 
#       you are sure you are not exposed to DNS spoofing attack.
#
#networkaddress.cache.ttl=-1 
 
# The Java-level namelookup cache policy for failed lookups:
#
# any negative value: cache forever
# any positive value: the number of seconds to cache negative lookup results
# zero: do not cache
#
# In some Microsoft Windows networking environments that employ
# the WINS name service in addition to DNS, name service lookups
# that fail may take a noticeably long time to return (approx. 5 seconds).
# For this reason the default caching policy is to maintain these
# results for 10 seconds. 
#
#
networkaddress.cache.negative.ttl=10

查看了jboss的run.sh脚本并未设置 java.security 相关参数，那我们的默认缓存时间应该是30 seconds

理论依据往往没有实验结果让人信服，于是又继续搜索相关的内容，终于在stackoverflow上找到了可以输出缓存内容的脚本。
我稍微修改了一下：

^?View Code JAVA

import java.lang.reflect.Field;
import java.net.InetAddress;
import java.net.UnknownHostException;
import java.util.ArrayList;
import java.util.Date;
import java.util.List;
import java.util.Map;
import java.text.SimpleDateFormat;
 
 
public class DNSCache {
  public static void main(String[] args) throws Exception {
    Date d = new Date();
    SimpleDateFormat sdf = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");
    InetAddress.getByName("www.google.com");
    try {
        InetAddress.getByName("nowhere.example.com");
    } catch (UnknownHostException e) {
 
    }
 
    System.out.println("current time:" + sdf.format(d));
    String addressCache = "addressCache";
    System.out.println(addressCache);
    printDNSCache(addressCache);
    String negativeCache = "negativeCache";
    System.out.println(negativeCache);
    printDNSCache(negativeCache);
  }
  private static void printDNSCache(String cacheName) throws Exception {
    Class<InetAddress> klass = InetAddress.class;
    Field acf = klass.getDeclaredField(cacheName);
    acf.setAccessible(true);
    Object addressCache = acf.get(null);
    Class cacheKlass = addressCache.getClass();
    Field cf = cacheKlass.getDeclaredField("cache");
    cf.setAccessible(true);
    Map<String, Object> cache = (Map<String, Object>) cf.get(addressCache);
    for (Map.Entry<String, Object> hi : cache.entrySet()) {
        Object cacheEntry = hi.getValue();
        Class cacheEntryKlass = cacheEntry.getClass();
        Field expf = cacheEntryKlass.getDeclaredField("expiration");
        expf.setAccessible(true);
        long expires = (Long) expf.get(cacheEntry);
 
        Field af = cacheEntryKlass.getDeclaredField("address");
        af.setAccessible(true);
        InetAddress[] addresses = (InetAddress[]) af.get(cacheEntry);
        List<String> ads = new ArrayList<String>(addresses.length);
        for (InetAddress address : addresses) {
            ads.add(address.getHostAddress());
        }
 
        System.out.println(hi.getKey() + " "+new Date(expires) +" " +ads);
    }
  }
}

编译 javac -Xlint:unchecked DNSCache.java
执行 java DNSCache 得到结果：

current time:2012-07-27 11:35:31
addressCache
0.0.0.0 Fri Jul 27 11:36:01 CST 2012 [0.0.0.0]
www.google.com Fri Jul 27 11:36:01 CST 2012 [74.125.71.105, 74.125.71.106, 74.125.71.147, 74.125.71.99, 74.125.71.103, 74.125.71.104]
negativeCache
nowhere.example.com Fri Jul 27 11:35:41 CST 2012 [0.0.0.0]

解析成功的域名www.google.com 缓存时间正好30 seconds
解析失败的域名nowhere.example.com 缓存时间正好10 seconds
与前面的理论完全对上，而且我们还看到对于多条A记录的域名它会全部缓存起来，并不是只缓存其中的一条A记录。
这里又引出了一个疑问：对于多条A记录是采用轮循还是什么策略使用呢？

我们可以修改脚本测试一下：

^?View Code JAVA

 
import java.lang.reflect.Field;
import java.net.InetAddress;
import java.net.UnknownHostException;
import java.util.ArrayList;
import java.util.Date;
import java.util.List;
import java.util.Map;
import java.text.SimpleDateFormat;
 
 
public class DNSCache {
  public static void main(String[] args) throws Exception {
    System.out.println("start loop\n\n");
    for(int i = 0; i < 30; ++i) {
        Date d = new Date();
        SimpleDateFormat sdf = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");
        System.out.println("current time:" + sdf.format(d));
        InetAddress addr1 = InetAddress.getByName("www.google.com");
        String addressCache = "addressCache";
        System.out.println(addressCache);
        printDNSCache(addressCache);
        System.out.println("getHostAddress:" + addr1.getHostAddress());
        System.out.println("*******************************************");
        System.out.println("\n");
        java.lang.Thread.sleep(10000);
    }
 
    System.out.println("end loop");
  }
 
 
  private static void printDNSCache(String cacheName) throws Exception {
    Class<InetAddress> klass = InetAddress.class;
    Field acf = klass.getDeclaredField(cacheName);
    acf.setAccessible(true);
    Object addressCache = acf.get(null);
    Class cacheKlass = addressCache.getClass();
    Field cf = cacheKlass.getDeclaredField("cache");
    cf.setAccessible(true);
    Map<String, Object> cache = (Map<String, Object>) cf.get(addressCache);
    for (Map.Entry<String, Object> hi : cache.entrySet()) {
        Object cacheEntry = hi.getValue();
        Class cacheEntryKlass = cacheEntry.getClass();
        Field expf = cacheEntryKlass.getDeclaredField("expiration");
        expf.setAccessible(true);
        long expires = (Long) expf.get(cacheEntry);
 
        Field af = cacheEntryKlass.getDeclaredField("address");
        af.setAccessible(true);
        InetAddress[] addresses = (InetAddress[]) af.get(cacheEntry);
        List<String> ads = new ArrayList<String>(addresses.length);
        for (InetAddress address : addresses) {
            ads.add(address.getHostAddress());
        }
 
        System.out.println(hi.getKey() + " "+new Date(expires) +" " +ads);
    }
  }
}

编译执行

^?View Code BASH

start loop
 
 
current time:2012-07-28 15:30:58
addressCache
0.0.0.0 Sat Jul 28 15:31:28 CST 2012 [0.0.0.0]
www.google.com Sat Jul 28 15:31:28 CST 2012 [74.125.71.103, 74.125.71.104, 74.125.71.105, 74.125.71.106, 74.125.71.147, 74.125.71.99]
getHostAddress:74.125.71.103
*******************************************
 
 
current time:2012-07-28 15:31:08
addressCache
0.0.0.0 Sat Jul 28 15:31:28 CST 2012 [0.0.0.0]
www.google.com Sat Jul 28 15:31:28 CST 2012 [74.125.71.103, 74.125.71.104, 74.125.71.105, 74.125.71.106, 74.125.71.147, 74.125.71.99]
getHostAddress:74.125.71.103
*******************************************
 
 
current time:2012-07-28 15:31:18
addressCache
0.0.0.0 Sat Jul 28 15:31:28 CST 2012 [0.0.0.0]
www.google.com Sat Jul 28 15:31:28 CST 2012 [74.125.71.103, 74.125.71.104, 74.125.71.105, 74.125.71.106, 74.125.71.147, 74.125.71.99]
getHostAddress:74.125.71.103
*******************************************
 
 
current time:2012-07-28 15:31:28
addressCache
www.google.com Sat Jul 28 15:31:58 CST 2012 [74.125.71.104, 74.125.71.105, 74.125.71.106, 74.125.71.147, 74.125.71.99, 74.125.71.103]
getHostAddress:74.125.71.104
*******************************************
#后面省略

结论：在缓存有效期内，取到的IP永远是缓存中全部A记录的第一条，并没有轮循之类的策略。
缓存失效之后重新进行DNS解析，因为每次域名解析返回的A记录顺序会发生变化(dig www.google.com测试可见)，所以缓存中的数据顺序也变了，取到的IP也变化。

当然最可靠的还是看下源码实现，有研究的同学请告诉我一下:)

最后附上几种修改缓存时间的方法：
1. jvm启动参数里面配置-Dsun.net.inetaddr.ttl=value
2. 修改 配置文件JAVA_HOME/jre/lib/security/java.security相应的参数networkaddress.cache.ttl=value
3. 代码里直接设置：java.security.Security.setProperty(”networkaddress.cache.ttl” , “value”);

参考资料：
http://docs.oracle.com/javase/6/docs/api/java/net/InetAddress.html
http://kenwublog.com/java-dns-cache-setting
http://stackoverflow.com/questions/1835421/java-dns-cache-viewer
http://docs.jboss.org/jbossas/docs/Server_Configuration_Guide/4/html/Security_on_JBoss-Running_JBoss_with_a_Java_2_security_manager.html

 

来源：飘零的代码 piao2010 ’s blog

评论

1 楼 oldrat 2015-04-14  

还给出了 给出“对于多条A记录是采用什么策略返回IP”的结论，赞！

写了操作Java DNS Cache的库 https://github.com/alibaba/java-dns-cache-manipulator ，方便用起来，支持 JDK 6 7 8。可以试试 ～ ：）