Connections get closed for no apparent reason but no exceptions are thrown

[Michael McGrady]

I can add that I have not even attempted to maximize performance as  
yet and I am getting amazing throughput involving very high levels of  
CPU usage for encryption, file writing, etc.  I will eventually get  
precise metrics that will be useful.  But, I now know that I will have  
no apparent reason ever to look beyond Netty for my performance  
requirements.  With double file writing linked synchronously to Netty,  
encryption, and significant application use which is also  
synchronously linked, I am getting about .28 millisecond response time  
per send for about 3,000 bytes average per send.  Thus, I am getting  
about 30,000 packets passed per 8 seconds.  This is with monitored  
throttling on the server reception side.

Mike

On Sep 9, 2009, at 10:57 PM, 이희승 (Trustin Lee) wrote:

> Hi Davis,
>
> On Tue, 8 Sep 2009 14:46:57 +0900 (KST)
> dralves <davidralves at gmail.com> wrote:
>>
>> Hi
>>
>> I'm using netty (3.1.2 GA) as the async protocol in a large cluster
>> application prototype I use for research.
>> Recently my client stopped being able to communicate with the server.
>>
>> Its strange because in my local computer everything works fine both
>> the client and the server show the connections as open and start to
>> communicate. When I deploy it on the cluster however (win 20008
>> server or ubuntu jaunty (clients)-> ubuntu jaunty (xen domU server))
>> the communication channel is open but is immediately closed
>> afterwards, and no exception is thrown as to why the connection was
>> closed (both on the client and the server).
>>
>> This only started to happen a little time ago, (the prototype was
>> undergoing big changes) but I can't seem to figure out why.
>>
>> Any ideas?
>
> If no exception is raised but connection is closed, it could mean that
> the connection has been closed normally IMHO.  It might be some
> firewall issue, but I'm not sure about your network configuration.
> Please let me know if you find something about this.
>
>> On another note how to make sure TRACE level logging is turned on in
>> netty? Or how to force netty to use Log4j? I have a uber jar with all
>> the dependencies that includes log4j but also sfl4j and adding the
>> line log4j.logger.org.jboss.nety=TRACE doesn't seem to do anything,
>> only the logs in my own hooks (like handlers) appear.
>
> Check org.jboss.netty.logging.
>
>> O another note before I had this problem I benchmarked netty in my
>> application against other frameworks (like mina) and its rocks. The
>> only problem I had before, which was latency, is solved now, i.e.
>> latency depends only on the frequency of flushes and is very stable,
>> and netty still achieves the highest throughputs.
>
> Happy to hear that!  It would be icing on the case if you could  
> publish
> your performance test result so that more people are convinced to
> use Netty. :)
>
> HTH,
> Trustin
>
> -- 
> Trustin Lee, http://gleamynode.net/
> _______________________________________________
> netty-users mailing list
> netty-users at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/netty-users

Mike McGrady
Principal Investigator AF081-028 AFRL SBIR
Senior Engineer
Topia Technology, Inc
1.253.720.3365
mmcgrady at topiatechnology.com