tomcat的一个常见错误

 

现象：

上次同事在热部署应用到liferay服务器上时，遇到了以下错误：

Sep 21, 2012 3:19:06 AM org.apache.catalina.startup.HostConfig deployDirectory

INFO: Deploying web application directory /home/portal/liferay-container-1.0.0-SNAPSHOT/standalone/liferay-portal/community-edition-liferay-portal-tomcat-6.1.0/liferay-portal-6.1.0-ce-ga1/tomcat-7.0.23/webapps/paas_integration_portlet

Sep 21, 2012 3:19:06 AM org.apache.catalina.core.StandardContext startInternal

SEVERE: Error listenerStart

Sep 21, 2012 3:19:06 AM org.apache.catalina.core.StandardContext startInternal

SEVERE: Context [/paas_integration_portlet] startup failed due to previous errors

Sep 21, 2012 3:19:06 AM org.apache.catalina.loader.WebappClassLoader checkThreadLocalMapForLeaks

SEVERE: The web application [/paas_integration_portlet] created a ThreadLocal with key of type [java.lang.ThreadLocal] (value [java.lang.ThreadLocal@3e043340]) and a value of type [java.lang.Class] (value [class oracle.sql.AnyDataFactory]) but failed to remove it when the web application was stopped. Threads are going to be renewed over time to try and avoid a probable memory leak.

Sep 21, 2012 3:19:06 AM org.apache.catalina.loader.WebappClassLoader checkThreadLocalMapForLeaks

SEVERE: The web application [/paas_integration_portlet] created a ThreadLocal with key of type [java.lang.ThreadLocal] (value [java.lang.ThreadLocal@5c0ce8f]) and a value of type [java.lang.Class] (value [class oracle.sql.TypeDescriptorFactory]) but failed to remove it when the web application was stopped. Threads are going to be renewed over time to try and avoid a probable memory leak.

Sep 21, 2012 3:19:06 AM org.apache.coyote.AbstractProtocol start

INFO: Starting ProtocolHandler ["http-bio-8080"]

Sep 21, 2012 3:19:06 AM org.apache.coyote.AbstractProtocol start

INFO: Starting ProtocolHandler ["ajp-bio-8009"]

Sep 21, 2012 3:19:06 AM org.apache.catalina.startup.Catalina start

INFO: Server startup in 37724 ms

 

基于我的以前经验，这是"类加载器泄露“的问题，具体细节参见我的以下邮件回复：

 

Reason:

 

This is very common in Tomcat .

 

We all know (1) ThreadLocal is used for holding variables per thread.

                         (2) Tomcat supports multi-threads.

                        (3) The default GC mechanism for tomcat is “object reference”

 

 

 

So the process is ,when we redeploy a portlet ,then the WebClassloader will load all the necessary classes and jars which required by this portlet ,and the threadlocal is used for identifying the thread . But when we destroy the portlet(redeploy or remove) ,though the portlet is removed ,but the threadlocal  value didn’t remove ,(which means the object reference number still >0) ,so the classes loaded by WebappClassLoader can’t not be gc , that’s a very famous topic  “Classloader leak”

 

My highlighted part is my understanding to this problem and it can explains what Judith has highlighted .

 

 

 

Solution:

 

“Classloader leak” issue happened since the first version of tomcat ,and it always a challenging work , as I know ,since tomcat 7 (because our Liferay 6.1 based on tomcat 7) has added some detection mechanism to detect the classloader leak issue ,that’s why it report “SEVERE”   ,if you familiar with tomcat 5 or 6 ,it just shutdown without any useful information.

 

Up to now just as I know ,we have no solution about it ,maybe we need to new version of tomcat to fix it .  If often meet with it ,I suggest we re-start the liferay.

 

 

 

Referenced Articles:

 

http://wiki.apache.org/tomcat/MemoryLeakProtection#customThreadLocal

https://issues.apache.org/bugzilla/show_bug.cgi?id=49159

http://www.javacodegeeks.com/2012/05/threading-stories-threadlocal-in-web.html

 

 

 

Charles