机一次cpu100%的线上故障
Posted zjhgx
tags:
篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了机一次cpu100%的线上故障相关的知识,希望对你有一定的参考价值。
某天发现线上crm机器cpu100%了,估计是哪里写了个死循环,用jstack看了下
at java.util.HashMap.hash(HashMap.java:351)
at java.util.HashMap.getEntry(HashMap.java:443)
at java.util.HashMap.containsKey(HashMap.java:434)
at ognl.OgnlContext.get(OgnlContext.java:491)
at com.opensymphony.xwork2.ActionContext.get(ActionContext.java:341)
at org.apache.struts2.ServletActionContext.getRequest(ServletActionContext.java:112)
at com.upg.ucars.framework.base.SessionTool.getAttribute(SessionTool.java:81)
at com.upg.ucars.framework.interceptor.AsyncInterceptor.actionIsLocked(AsyncInterceptor.java:71)
at com.upg.ucars.framework.interceptor.AsyncInterceptor.intercept(AsyncInterceptor.java:38)
at com.opensymphony.xwork2.DefaultActionInvocation.invoke(DefaultActionInvocation.java:246)
at com.upg.ucars.framework.interceptor.AuthorizationInterceptor.intercept(AuthorizationInterceptor.java:83)
at com.opensymphony.xwork2.DefaultActionInvocation.invoke(DefaultActionInvocation.java:246)
at com.opensymphony.xwork2.interceptor.ConversionErrorInterceptor.intercept(ConversionErrorInterceptor.java:138)
at com.opensymphony.xwork2.DefaultActionInvocation.invoke(DefaultActionInvocation.java:246)
at com.opensymphony.xwork2.interceptor.ParametersInterceptor.doIntercept(ParametersInterceptor.java:249)
at com.opensymphony.xwork2.interceptor.MethodFilterInterceptor.intercept(MethodFilterInterceptor.java:98)
at com.opensymphony.xwork2.DefaultActionInvocation.invoke(DefaultActionInvocation.java:246)
at com.opensymphony.xwork2.interceptor.ParametersInterceptor.doIntercept(ParametersInterceptor.java:249)
at com.opensymphony.xwork2.interceptor.MethodFilterInterceptor.intercept(MethodFilterInterceptor.java:98)
at com.opensymphony.xwork2.DefaultActionInvocation.invoke(DefaultActionInvocation.java:246)
at com.opensymphony.xwork2.interceptor.StaticParametersInterceptor.intercept(StaticParametersInterceptor.java:191)
at com.opensymphony.xwork2.DefaultActionInvocation.invoke(DefaultActionInvocation.java:246)
at org.apache.struts2.interceptor.FileUploadInterceptor.intercept(FileUploadInterceptor.java:252)
at com.opensymphony.xwork2.DefaultActionInvocation.invoke(DefaultActionInvocation.java:246)
at com.opensymphony.xwork2.interceptor.ChainingInterceptor.intercept(ChainingInterceptor.java:145)
at com.opensymphony.xwork2.DefaultActionInvocation.invoke(DefaultActionInvocation.java:246)
at com.opensymphony.xwork2.interceptor.I18nInterceptor.intercept(I18nInterceptor.java:139)
at com.opensymphony.xwork2.DefaultActionInvocation.invoke(DefaultActionInvocation.java:246)
at org.apache.struts2.interceptor.ServletConfigInterceptor.intercept(ServletConfigInterceptor.java:164)
定位到 at com.upg.ucars.framework.interceptor.AsyncInterceptor.actionIsLocked(AsyncInterceptor.java:71)这一句
相关的代码如下:
public String intercept(ActionInvocation arg0) throws Exception {
try{
if(SessionTool.getLocale()==null){
SessionTool.setDefaultLocale();
}
String actionName=arg0.getProxy().getActionName().toUpperCase();
if("LOGIN".equals(actionName)){
unlockAction();
}
boolean lock=true;
if(isUncontrolled(arg0))
return arg0.invoke();
else{
while(lock){
lock=actionIsLocked();
if(!lock){
lockAction();
String target=arg0.invoke();
unlockAction();
return target;
}
}
}
}catch(Exception e){
unlockAction();
throw e;
}
return "";
}
因为这里有个while循环,估计就是在这里死循环了
/**
* 判断action是否已加锁
* @return
*/
private boolean actionIsLocked(){
String flag=(String)SessionTool.getAttribute(SessionKeyConst.ASYNC_ACTION_LOCK);
if(flag==null)
return false;
return flag.equals("true");
}
/**
* 为action加锁
*/
private void lockAction(){
SessionTool.setAttribute(SessionKeyConst.ASYNC_ACTION_LOCK, "true");
}
/**
* 为action解锁
*/
private void unlockAction(){
SessionTool.setAttribute(SessionKeyConst.ASYNC_ACTION_LOCK, "false");
}
}
/**
* (获取session中存储的数据对象)
* @param key 数据对象在session中的唯一标识
* @return 存储在session中的数据对象
* @date
*/
public static Object getAttribute(String key){
Object ret=null;
try{
HttpServletRequest request=ServletActionContext.getRequest();
if(request!=null){
ret=request.getSession().getAttribute(key);
}
}catch(NullPointerException e){
return ret;
}
return ret;
}
这里的加锁和解锁就是在session里把SessionKeyConst.ASYNC_ACTION_LOCK这个属性设成true和false,但HttpSession这个类对于同一个用户的并发操作是非线程安全的,有可能一个在解锁同时另一个在加锁,一个线程在加锁后变成他true,此时另一个线程刚好调用actionIsLocked返回true则就进入死循环了。去线上一看,果然有同一个用户
以上是关于机一次cpu100%的线上故障的主要内容,如果未能解决你的问题,请参考以下文章
记一次线上故障--HashMap在多线程条件下运行造成CPU 100%