Synchronized is important to update the thread caches. More so when different sessions are bound to different threads. On Nov 21, 2013 10:45 AM, "Alexandros Touloupis" <atoulou...@gmail.com> wrote:
> Thanks for the reply! > > Yes I agree that synchronizing to the whole IoSession instance each time > for writing, is brutal. I am thinking on including a writing lock as a > session attribute in each session that will be initialized on > sessionCreated events. > > Cheers! > Alex > > > On Thu, Nov 21, 2013 at 4:28 PM, Jon V. <sybersn...@gmail.com> wrote: > > > I would imagine this to be true. The normal use case is request and > > response. Additional synchronization makes the code run slower in that > > scenario. I would suggest synchronizing critical paths. > > > > I have to synchronize write paths for the same reason where one input may > > write to dozens of other sockets. > > On Nov 21, 2013 10:23 AM, "Alexandros Touloupis" <atoulou...@gmail.com> > > wrote: > > > > > Hello all, > > > > > > I am using mina 2.0.8 since I needed a patch included in this release. > > > > > > I face the following problem: > > > > > > The concept of the application is to "multiplex" inbound to outbound > > > connections in MINA. This means that there might be 2 incoming > > connections > > > that are multiplexed in one outbound connection. During my tests i > > randomly > > > see this exception: > > > > > > java.nio.InvalidMarkException > > > at java.nio.Buffer.reset(Buffer.java:298) ~[?:1.7.0_40] > > > at > > > > > > > > > org.apache.mina.core.buffer.AbstractIoBuffer.reset(AbstractIoBuffer.java:414) > > > ~[mina-core-2.0.8-PATCHED.jar:?] > > > at > > > > > > > > > org.apache.mina.core.polling.AbstractPollingIoProcessor.writeBuffer(AbstractPollingIoProcessor.java:926) > > > ~[mina-core-2.0.8-SNAPSHOT.jar:?] > > > at > > > > > > > > > org.apache.mina.core.polling.AbstractPollingIoProcessor.flushNow(AbstractPollingIoProcessor.java:844) > > > [mina-core-2.0.8-SNAPSHOT.jar:?] > > > at > > > > > > > > > org.apache.mina.core.polling.AbstractPollingIoProcessor.flush(AbstractPollingIoProcessor.java:772) > > > [mina-core-2.0.8-SNAPSHOT.jar:?] > > > at > > > > > > > > > org.apache.mina.core.polling.AbstractPollingIoProcessor.access$700(AbstractPollingIoProcessor.java:67) > > > [mina-core-2.0.8-SNAPSHOT.jar:?] > > > at > > > > > > > > > org.apache.mina.core.polling.AbstractPollingIoProcessor$Processor.run(AbstractPollingIoProcessor.java:1112) > > > [mina-core-2.0.8-SNAPSHOT.jar:?] > > > at > > > > > > > > > org.apache.mina.util.NamePreservingRunnable.run(NamePreservingRunnable.java:64) > > > [mina-core-2.0.8-SNAPSHOT.jar:?] > > > at > > > > > > > > > java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145) > > > [?:1.7.0_40] > > > at > > > > > > > > > java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615) > > > [?:1.7.0_40] > > > at java.lang.Thread.run(Thread.java:724) [?:1.7.0_40] > > > > > > After some investigation i noticed that this happened when 2 threads > > > decided to write to the same IoSession a message the same time. My > > > assumption is that the #IoSession.write method is not thread safe, > thus i > > > synchronized the session instance before executing the write method in > my > > > handlers and up until now i haven't faced the same issue. Is my > > assumption > > > valid? > > > > > > Thank you in advance, > > > Alex > > > > > >
