[Jetty-support] Re: Thread Dump ... @ 99% CPU

[Greg W. <gr...@mo...>]

Alan,

as you say, there is nothing terribly obvious in that thread dump.

When the server goes 99% I assume it is still serving requests?
Judging by the dump it has some requests in progress.

Next time this happens, can you take two thread dumps a few
minutes apart - that will show us if the same thread is in
the same place.

My current suspect is BoundedThreadPool0-31 as that is in 
an area that could loop without actually hitting a real IO
exception when the connection times out.   But code inspection
is showing nothing obvious.  Note that your line numbers do
not match mine exactly and I don't think this class has changed
for a while (but svn has gone silent on me... so I can't check).

But thanks for sticking with this!

cheers




Alan Williamson wrote:
> I have managed to capture a thread dump of when the system goes into a
> 99% CPU chew
> 
> Thread dump here:
>   http://whiteboard.blog-city.org/index.cfm?boardname=jettythreaddump
> 
> 
> There is nothing there that is terribly obvious.  Although this is
> another one that shows a high number of BoundedThreadPool0-32 instances
> that are just not cleaned up.
> 
> I spoke to Kirk Pepperdine of JavaPerformance.com about this, and he
> asked if Jetty uses the same pool libraries that Tomcat utilises?  He
> noted that Tomcat has the same behaviour that it doesn't release unused
> threads back.
> 
> The Java GC isn't doing anything either, so one of those threads is
> chewing up the processor
> 
> Thoughts?
>