未同步hashMap于并发环境cpu巨高

 

 系统中 一小疏漏，使用 LRUMap作缓存，其继承至abstractHashMap, 因hashmap isn't thead-safe.

并发量大时, 导致客户一系统全部hung住. 只能重启. 

后查was也发生过类似问题. 

  WebSphere Portal 6.1 and 7 is running on WebSphere Application Server 7 which is based on Java 6. The Java 6 HashMap which is not thread-safe is less tolerant to concurrent access under load. This causes the threads accessing the HashMap concurrently to hang and the WebSphere Portal process to incur a higher than normal CPU usage. In Many Cases the hung threads can lead to near 100% CPU usage by the WebSphere Portal process.

 

http://www-01.ibm.com/support/docview.wss?uid=swg21427374