[ROOT] Re: root problem (fwd)

From: Fons Rademakers (Fons.Rademakers@cern.ch)
Date: Fri Feb 14 2003 - 19:56:06 MET


Hi Steven,

 could it be that on the server side the socket is not properly closed?
I use TSocket to a httpd daemon and close it via standard
TSocket::Close() and there are no problems.

Cheers, Fons.


On Thu, 2003-02-13 at 21:49, stevenkj wrote:
> Hello,
> 
> 	I'm having real trouble right now with getting socket
> communication to work in root (Version   3.03/07    14 August 2002).
> I'm opening  tscoket in a loop
> 
>     TSocket *s = new TSocket("localhost", 9090);
> 
>     then transfer histograms across the tsocket,
> 
> s->Recv(mess_in);
>   if (mess_in){
>     if (mess_in->What() == kMESS_OBJECT)
>       HTrksPSec = (TH1F *)mess_in->ReadObject(mess_in->GetClass());
>     else {
>       cout << "***Unexpected message***" << endl;
>       if (mess_in->What() == kMESS_STRING)
> 	{ char errstr[64];  mess_in->ReadString(errstr, 64);
> printf("%s\n", errstr); }
>     }
>   }
> I then close the tsocket before continuing in another loop iteration.
> s->Close()
> 
> ps. I tried s->Close("force"); which didn't work either.
> 
> 	The problem is that root doesn't seem to close the socket properly
> which is causing serious problems with our online monitoring (see mail
> below) & is starting to hurt our trigger comissioning. Has anyone ever
> seen this before, is it a known problem, is there a solution ?
> 
> 				Kyle.
> 
> 
> ---------- Forwarded message ----------
> Date: Fri, 07 Feb 2003 17:12:19 -0600 (CST)
> From: Stu Fuess <fuess@fnal.gov>
> To: Kyle Stevenson <stevenkj@fnal.gov>
> Subject: root problem
> 
> We looked at file descriptor usage:
> 
> [root@d0ol43 ~]# ls -1 /proc/23124/fd | wc -l
>    1016
> [root@d0ol43 ~]# ls -1 /proc/23124/fd | wc -l
>    1018
> [root@d0ol43 ~]# ls -1 /proc/23124/fd | wc -l
>    1020
> [root@d0ol43 ~]# ls -1 /proc/23124/fd | wc -l
>    1024
> [root@d0ol43 ~]# ls -1 /proc/23124/fd | wc -l
>    1024
> [root@d0ol43 ~]# ls -1 /proc/23124/fd | wc -l
>    1024
> 
> It hits the kernel limit of 1024 files per process max.
> 
> where 23124 was the pid of
> ../../bin/Linux2.4-KCC_4_0/l1CTT_examine_x -rcp  ../rcp/runol_vsvx_framework.rcp
> 
> 
> see from netstat -a that CLOSE_WAIT sockets never go away:
> 
> tcp        1      0 localhost.localdom:9090 localhost.localdo:39646 CLOSE_WAIT
> tcp        1      0 localhost.localdom:9090 localhost.localdo:38110 CLOSE_WAIT
> tcp        1      0 localhost.localdom:9090 localhost.localdo:40157 CLOSE_WAIT
> tcp        1      0 localhost.localdom:9090 localhost.localdo:39133 CLOSE_WAIT
> tcp        1      0 localhost.localdom:9090 localhost.localdo:39389 CLOSE_WAIT
> tcp        1      0 localhost.localdom:9090 localhost.localdo:38365 CLOSE_WAIT
> tcp        1      0 localhost.localdom:9090 localhost.localdo:41180 CLOSE_WAIT
> tcp        1      0 localhost.localdom:9090 localhost.localdo:38620 CLOSE_WAIT
> 
> 
> 
-- 
Org:    CERN, European Laboratory for Particle Physics.
Mail:   1211 Geneve 23, Switzerland
E-Mail: Fons.Rademakers@cern.ch              Phone: +41 22 7679248
WWW:    http://root.cern.ch/~rdm/            Fax:   +41 22 7679480



This archive was generated by hypermail 2b29 : Thu Jan 01 2004 - 17:50:09 MET