tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mark Thomas <>
Subject Re: Question on bug# 53267 in tomcat 7.0.47
Date Wed, 11 Dec 2013 19:52:01 GMT
On 11/12/2013 19:36, bharath chinnadurai maharajan wrote:
> Hello,
> This is my first post to the tomcat forum, so please excuse of any errors.
> I am looking at a fix for one of the bug (bug id#53267) in Catalina. Ref -
> The bug is in RESOLVED FIXED state.
> As per this bug, the one hr time interval (when a GC event is kicked off)
> was changed to Long.MAX ( a very long time).

No it wasn't. Long.MAX_VALUE has a special meaning. It was changed to

> It is also mentioned the value
> is configurable using a system property -
> org.apache.catalina.core.jreMemoryLeakPreventionGCDaemonPeriod.

It is also mention that the addition of the system property was reverted.

> When I tried setting this property to 3600000, assuming a full GC will be
> kicked every hour it did not happen. Curious, I looked into the code for
> version 7.0.47, that config variable was not used in
> org.apache.catalina.core.JreMemoryLeakPreventionListener.

As intended.

> Code snippet from JreMemoryLeakPreventionListener in tomcat version 7.0.47 :
>  Class<?> clazz = Class.forName("sun.misc.GC");
>                         Method method = clazz.getDeclaredMethod(
>                                 "requestLatency",
>                                 new Class[] {long.class});
>                         method.invoke(null, Long.valueOf(Long.MAX_VALUE -
> 1));
> Instead of using the Long.MAX_VALUE-1 by default, it should try and resolve
> if from the aforementioned system property.

No it should not. This code does not exist to enable you to set a
regular interval for GC. If you want to do that you need to write your
own code to do so.

> Am I wrong to assume the config variable will be present in 7.0.47?


> Thanks for your time, and my sincere apologies if I had wasted it.

I suggest you read the thread linked from comment 11 and look at commits
mentioned in the report.


To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message