HP3000-L Archives

December 1996, Week 5

HP3000-L@RAVEN.UTC.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Jeff Kell <[log in to unmask]>
Reply To:
Jeff Kell <[log in to unmask]>
Date:
Mon, 30 Dec 1996 19:50:18 -0500
Content-Type:
text/plain
Parts/Attachments:
text/plain (31 lines)
Leon Degeling wrote:

> I use dce/3000 to communicate from my 3000/987 (client) application to
> my 9000 (server) application (rpc calls).  If more than 43 sessions
> try to bind to the dce-servers the programme aborts saying: can't get
> binding: error with socket. Since there are over 200 other clients
> connecting to the server I'm sure the problem is somewhere on my 3000
>
> Someone from hprc suggested that I might be running out of sockets, so
> I started NETTOOL.NET and found out that i was using abort 340 sockets
> on my system.  [...]  Next i started nmconfig and found a [max] number
> of connections (512) somewhere under tcp (unguided). Now I figure I'm
> not running out of sockets. But am I making any sense?  And can anyone
> think of some limit concerning sockets/dce which I am running into.

Two problems come to mind (boy! deja-vu twice tonight!)...

There was a problem in 5.0 [pull?] where BSD socket code could leave
sockets in a closed yet "not quite free" state.  NETTOOL would report a
lower number than were actually "occupied" by these hung sockets.
Restarting the network would release some of the hung sockets.

There was another problem in 5.5 where inbound buffers were being
consumed.  This caused weird problems (halted network services, etc) but
could be avoided by restarting any jobs/sessions holding these buffers;
the most prevalent report was on NCSA httpd.

Perhaps one of these two?  (both have been fixed)

Jeff Kell <[log in to unmask]>

ATOM RSS1 RSS2