基于Java IO 序列化方案的memcached-session-manager多memcached节点配置

session的序列化方案官方推荐的有4种

  1. java serialization
  2. msm-kryo-serializer
  3. msm-javolution-serializer
  4. msm-xstream-serializer

关于这几种,官方也给出了比较:

  • Java serialization is very robust and a proven technology. The biggest disadvantage IMHO is that different class versions cannot be handled.
  • Kryo is an extremely fast binary serialization library. In the popular thrift-protobuf-compare benchmark it's one of the fastest serialization toolkits - and it differs from the fastest in that it does NOT need a schema definition of serialized data, which is a requirement for serialization arbitrary session data. A disadvantage of using kryo based serialization is that it's binary - you just cannot look how the serialized object graph looks like. This is my favorite serialization strategy, just because of its great performance.
  • Javolution is a very good and fast xml binding toolkit. The reflection part is written by me and adds the bits that are actually binding POJOs to xml. It is covered well with unit tests, however I cannot guarantee that there's no issue left to solve (actually this serialization strategy was in use in my own projects, now replaced by kryo based serialization).
  • XStream based serialization should be very robust as this is an often used java object binding library. The biggest disadvantage IMHO is the relatively bad performance.

要注意的是javolution是唯一支持copyCollectionsForSerialization="true"时对能对线程不安全的collection进行序列化特性的,其他的方案要对序列化线程不安全的collection时特别处理,性能最好的序列化方案是Kryo。

网上有一些例子都是关于Javolution的,我个人倾向于使用Kryo,至于序列化出现的问题还可以自己解决,不过先使用java IO来配置,先用起来,以后再慢慢优化,换成Kryo。

直接在CATALINAHOME/lib/下添加memcached−2.5.jar和memcached−session−manager−1.4.0.jar ,然后对CATALINA_HOME/conf/server.xml修改相应的配置

复制代码
   <Context path="/webapp" docBase="D:\webapp\WebRoot" reloadable="false">  
     <Manager   
        className= "de.javakaffee.web.msm.MemcachedBackupSessionManager"  
        memcachedNodes= "n1:192.168.112.1:11211,n2:192.168.112.2:11211"   
        sticky="false"   
        lockingMode="auto"   
        requestUriIgnorePattern= ".*\.(png|gif|jpg|css|js)$"    
        sessionBackupAsync= "false"   
        sessionBackupTimeout= "0"   
        memcachedProtocol="binary"  
        transcoderFactoryClass= "de.javakaffee.web.msm.JavaSerializationTranscoderFactory"    
     />  
   </Context>  

由于项目对session 管理模式是non-sticky的,所以不配置failoverNodes,任何一个web工程对session的修改要及时更新到memcache上,所以sessionBackupTimeout为0(不同的应用有不同的配置,这样配置是符合我现在这个项目要求的)。

相关推荐
树獭非懒3 小时前
AI大模型小白手册|Embedding 与向量数据库
后端·python·llm
NE_STOP4 小时前
MyBatis-配置文件解读及MyBatis为何不用编写Mapper接口的实现类
java
唐叔在学习6 小时前
就算没有服务器,我照样能够同步数据
后端·python·程序员
曲幽8 小时前
FastAPI流式输出实战与避坑指南:让AI像人一样“边想边说”
python·ai·fastapi·web·stream·chat·async·generator·ollama
Flittly8 小时前
【从零手写 AI Agent:learn-claude-code 项目实战笔记】(1)The Agent Loop (智能体循环)
python·agent
后端AI实验室9 小时前
用AI写代码,我差点把漏洞发上线:血泪总结的10个教训
java·ai
vivo互联网技术10 小时前
ICLR2026 | 视频虚化新突破!Any-to-Bokeh 一键生成电影感连贯效果
人工智能·python·深度学习
程序员清风11 小时前
小红书二面:Spring Boot的单例模式是如何实现的?
java·后端·面试
belhomme11 小时前
(面试题)Redis实现 IP 维度滑动窗口限流实践
java·面试