On 7/28/07, Gaston Dombiak <[EMAIL PROTECTED]> wrote: > > Hey Trustin, > > I noticed many times (specially when behind a NAT device) that sometimes > when a socket connection was lost/closed it's like the TCP layer does not > inform the JVM about the event. So the JVM thinks that the socket is fine > and even after writing some data (before the write queue gets full) the JVM > will not get any IO exception so it will still think that the socket is > alive. The only workarounds we applied (at the application level) were 1) > add heartbeats that will eventually (hopefully) will fill up the write queue > and get an IO exception or 2) close socket that have not "sent" any data to > the server. I think that #1 is not 100% safe so #2 will eventually close the > "assumed" dead connection.
Will setWriteTimeout solve uncaught IO exception problem? Or a setIdleTime for Read and Write and send ping-pong? Regards, > > -- Gato > > ----- Original Message ----- > From: "Trustin Lee" <[EMAIL PROTECTED]> > To: dev@mina.apache.org > Sent: Friday, July 27, 2007 10:11:14 AM (GMT-0800) America/Los_Angeles > Subject: Re: Out Of Memory Problem Again > > On 7/28/07, mat <[EMAIL PROTECTED]> wrote: > > After further research, I found out if one client disconnection was not > > caught and session.write() kept got called. Which causes the OOM? Is > that > > possible? > > Yes. You have to make sure if the connection is alive and write > operation was successful. You can check if the connection is alive by > calling IoSession.isClosing(), and WriteFuture.isWritten() for write > operations. Should that be session.isConnected()? However, if disconnection doesn't get caught, session.isConnected() will never work since you never do session.close(), right? SetWriteTimeout works for this issue, right? Why WriteFuture.isWritten()?? HTH, > Trustin > -- > what we call human nature is actually human habit > -- > http://gleamynode.net/ > -- > PGP Key ID: 0x0255ECA6 > >
