[jboss-user] [Remoting] - Re: Socket Read Timeout on stressing calls for validating pr

ron.sigal@jboss.com do-not-reply at jboss.com
Sun Jan 4 01:54:33 EST 2009


Hi Vinay,

You have described problems on this thread and two other threads, which is making it hard for me to get a unified picture of what's going on.  To try to clarify things, I'm consolidating the discussions on this thread.  Here, I believe, are the phenomena you have experienced:

1. This thread: 

"vink" wrote : 
  | On increasing timeout value of ejb3 connector, this exception is not coming. But, on running my stress tests, my application is throwing java.lang.OutOfMemoryError: unable to create new native thread 
  | 

Is it on the client side or the server side where you can't create a new thread?  I believe the thread dumps you've shown me are from the server side, and I don't see anything particularly unusual.

"vink" wrote : 
  | And, number of such sockets increases as time passes by & it finally results into the following exception; 
  | ...
  |  java.net.BindException: Address already in use: connect 
  | ...
  | 

If you're running on a Windows machine, you could be running up against a configurable limit Windows puts on the number of open sockets.  For example, the parameter "MaxUserPort", which defaults to 5000, limits the user ports to values between 1024 and 5000.  You can read more at, for example, "Microsoft Windows 2000 TCP/IP Implementation Details" - http://technet.microsoft.com/en-us/library/bb726981.aspx#EDAA.

2. "JBREM-877: New Socket Connection is being Created for Every" - http://www.jboss.com/index.html?module=bb&op=viewtopic&t=126382&start=10

"vink" wrote : 
  | On using invokerDestructionDelay in ejb3 connector configuration, my system goes slow & slow and run out of resources(native threads) in few minutes. 
  | 

Again, client side or server side?

"vink" wrote : 
  | When removing this attribute, the system don't run out of resources, but many sockets are visible in TIME_WAIT state. As a result of this, my system run out of sockets & after a while I get sporadic disconnects.
  | 

Client side or server side?  And, could you describe "sporadic disconnects" more precisely?

2. "Stream Corrupted Exception on Server side" - http://www.jboss.com/index.html?module=bb&op=viewtopic&t=147905

"vink" wrote : 
  | After several hours of running I'm getting following exception
  | ...
  | Caused by: java.io.StreamCorruptedException: invalid stream header: 77011679 
  | 

Clearly, this exception is occurring on the server side, but the actual problem might be occurring on the client side.  Do you see anything unusual happening on the client side when this exception is thrown?


===============================================
I have a couple of general questions.

1. Am I right in inferring that you're running a stress test and you're seeing performance deteriorating on the client side?

2. Can you create a simple version of your application that demonstrates the same problems and that I can run for myself?

Thanks,
Ron

View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4199374#4199374

Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=4199374



More information about the jboss-user mailing list