Re: new : iperf servers that stay alive


A supported fix for the problem of "pthread_create failed" after handling 
a lot of clients, is to add the following line:

pthread_detach( mTID );

immediately following line 118 of lib/Thread.cpp which is:

FAIL( err != 0, "pthread_create" );

The other "problems/bugs" were caused by Marc's changes (and possibly 
fixed as well).

Kevin

On Wed, 22 Jan 2003 Marc.Herbert --at-- ens-lyon.fr wrote:

> 
> On most systems, iperf servers had a limited life expectancy. I think
> you can bring down most TCP iperf servers using such a loop :
> 
> n=1; while (./iperf -t 1 -c server.to.crash -P 50 ) ; do  printf "\n\nn $n\n\n\n"; let n=$n+1; date; done
> 
> 
> The heavily commented patch "iperf165_thread.patch"
>    <http://marc.herbert.free.fr/TCP/iperf/>
> hopefully solves the issue. Any feedback on it is highly welcome
> (further test results, congratulations, insults,...) It was tested on
> debian woody  linux 2.4.18 and Solaris 2.6.
> 
> 
> The few UDP tests I performed so far seem to show that this patch also
> prevent UDP servers from crashing, but I would not guarantee it; fully
> understanding the behaviour of the iperf threads in case of
> unreliability and timeouts is much more complex than in the case of TCP.
> 
> 
> 
> Sincerely,
> 
> 



Other Mailing lists | Author Index | Date Index | Subject Index | Thread Index