hc-httpclient-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Oleg Kalnichevski <ol...@apache.org>
Subject Re: outputReady method in IOEventDispatcher
Date Thu, 27 Mar 2014 09:19:37 GMT
On Thu, 2014-03-27 at 14:13 +0530, Sajith Dilshan wrote:
> Hi Oleg,
> 
> Thank you for the reply. But I'm still a bit confused about the answer. So
> let me rephrase my question. When a worker thread enters
> outputReady(IOSession session) method with that session object, does that
> worker thread acquires a lock on  OP_WRITE operation for that particular
> session object so that any other worker thread can't perform OP_WRITE
> operation on that session object?
> 

Only I/O event dispatch threads managed by the I/O reactor should ever
call IOEventDispatch#outputReady. It is also guaranteed by the framework
that only one I/O dispatch thread can interact with the same I/O session
at a time.

If your application create additional worker threads those threads
should never call IOEventDispatch#outputReady. They should only interact
with IOSession, update its state but let the I/O dispatch threads do the
actual reading and writing.

> As an example, assume there are two worker threads, Thread_A and Thread_B.
> Further, there is a IOSession object session_123 with its event set as
> OP_WRITE. Both threads, Thread_A and Thread_B see that OP_WRITE event is
> set for session_123 and can the both threads (Thread_A, Thread_B) enter the
> outputReady(IOSession session) method at the same time with the session_123
> being the parameter passed to that method and execute the method's body?
> 

See above. Worker threads should update IOSession's state (such as
shared I/O buffer) but the actual writing should be performed by the I/O
dispatch thread.

There is a section on asynchronous I/O control for HTTP connections you
may find useful:

http://hc.apache.org/httpcomponents-core-4.3.x/tutorial/html/nio.html#d5e614

Oleg

> 
> On Thu, Mar 27, 2014 at 1:52 PM, Oleg Kalnichevski <olegk@apache.org> wrote:
> 
> > On Wed, 2014-03-26 at 23:00 +0530, Sajith Dilshan wrote:
> > > Hi everyone,
> > >
> > > When I set the OP_WRITE event on a particular instance of
> > > org.apache.http.nio.reactor.IOSession, a worker thread will pick up that
> > > IOSession and will execute the outputReady(IOSession session) method of
> > the
> > > respective org.apache.http.nio.reactor.IOEventDispatch implementation.
> > >
> > > My question is that, is it possible for more than one thread to execute
> > > outputReady(IOSession session) method for the same instance of IOSession
> > at
> > > the same time?
> >
> > it is possible on multi CPU core systems.
> >
> > > Does a worker thread acquires a lock before executing
> > > outputReady(IOSession session) method for a particular IOSession
> > instance?
> > >
> >
> > Yes, it does. Access to IOSession's internal structures is synchronized.
> > IOSession-s are fully thread safe.
> >
> > Oleg
> >
> >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: httpclient-users-unsubscribe@hc.apache.org
> > For additional commands, e-mail: httpclient-users-help@hc.apache.org
> >
> >



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


Mime
View raw message