hc-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ortwin Glück <ortwin.glu...@nose.ch>
Subject Re: ExecutorService, Future, and HttpClient with a MultiThreadedHttpConnectionManager
Date Thu, 31 Mar 2005 15:55:14 GMT

Kent Gibson wrote:
> thanks for the advice. 

You should turn on DEBUG level logging on the 
MultiThreadedHttpConnectionManager class:


It should give you some hints about what is happening.

Any chance you can debug or profile your application?

> hmm explicit gc calls. In an effort to avoid these I
> set max and min heap to the same and use -xincg. But I
> will try explicit calls.

True, explicit GC calls are normally bollocks. I know. But it's worth a 
try. I don't know why you selected the incremental GC. Have a look at
which may help you selecting the right one.

> File descriptors, i am running under windows, but
> where could I check this problem?

I don't know much about Windows (what was the name of that vendor again? 
just kidding...) as I primarily use Linux for real systems, and there we 
have lsof. I am not too sure if Windows handles sockets as files (like 
Unix does) at all. I also don't know if Windows has something like per 
user resource limits like Unix has (ulimit) and that could limit the 
number of file descriptors a specific user can use.

> I am using a final but I will have another look. There
> should be no problem though releasing an already
> released connection no?

You should not do that. The API Docs make no promise that this is 
allowed. Especially in a mulithreaded environment this could put a 
connection back to the pool although it is already used by another thead 
again! The consequences would be disasterous. We should address this 
issue in the 4.0 API actually to make a "double free" not suck.

Hope that helps

Ortwin Glück

To unsubscribe, e-mail: httpclient-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: httpclient-dev-help@jakarta.apache.org

View raw message