[olug] A TCP/IP mystery

[Brian Roberson]

First off, you need to determine what is "blocking" - e.g. is it the network stack or
the application itself.


For a quick look at the network, I would fire up tcpdump for a while and import your log file 
into Ethereal... I mean WireShark...  If the problem is at the tcp/network level, you will 
see items like missed ack's and resends and other items that will be flagged red :-). 

You can use "tcpdump -w {filename}" for this 


To look at the application, I would need to understand how it is interacting with the 
socket, e.g. if it is a Perl application, are you using the IO::Socket module, or 
interacting with the socket directly, or are you using Net:SSLeay etc..etc..etc..

At the application level, there are alot of options for debuging, most notably using
Errno functionality if using perl and getting the raw socket error conditions; Such 
as EWOULDBLOCK if you set the socket to non-blocking, etc..






Or Wed, Mar 12, 2008 at 06:15:59PM -0500, Dave Thacker wrote:

> We have a process at work that sends data from a windows machine (owned by 
> vendor S) through an encrypted TCP tunnel to a linux machine.   The lash up 
> is not performing well and finger-pointing abounds.   Currently, Vendor S is 
> telling us that the Linux machine is "blocking" when they attempt to send a 
> large volume of transactions to it. 
> 
> I'm not seeing errors in the the stunnel log, nor am I showing errors in the 
> basic ifconfig command.  This is an SLES 10 box, so I have the usual array of 
> linux implements available, and I've installed iptraf.   If I was to start 
> monitoring with those tools. what would a "block" look like?  
> 
> TIA
> Dave
> 
> _______________________________________________
> OLUG mailing list
> OLUG at olug.org
> http://lists.olug.org/mailman/listinfo/olug