Memory leak when using GoAhead webserver.

Jake Janovetz janovetz at uiuc.edu
Mon Jan 22 19:46:27 UTC 2001


It may be the ethernet stack.  My terminology is going to be
whack, but I'll try.  When a TCP connection is created, used,
then subsequently closed, the something-or-other stays in the 
system (call tcpstat to see it) for a while.  After a timeout
period, the something-or-other is officially destroyed.

   Jake


On Mon, Jan 22, 2001 at 04:29:21PM -0000, Nick.SIMON at syntegra.bt.co.uk wrote:
> Lies, all lies.
> 
> On further investigation I don't seem to have a memory leak at all.
> 
> When a page is "got" from the webserver, a certain amount of memory - 488
> bytes or thereabouts - is allocated and not released until 60 seconds later.
> This gave the appearance of a leak, of course.
> 
> If anyone knows what part of the system is playing this prank, I'd love to
> know, just for academic interest.  I really can't justify spending time
> looking for it.
> 
> 
> 
> -- Nick (madder than a sackful of itchy ferrets) Simon 
> 
> > 
> > Nick.SIMON at syntegra.bt.co.uk wrote:
> > > 
> > > Hi,
> > > 
> > > I'm experiencing loss of heap when retrieving pages from 
> > webserver, about
> > > 500 bytes a time.  As the NT test version also has goahead, 
> > and doesn't lose
> > > heap, I suspect IMFS (where html files are kept) or 
> > possibly sockets.  I'll
> > > do some tests to isolate this when I can get to a target but, in the
> > > meantime, has anyone any knowledge of this?
> > > 
> > > I'm using 4.5.0 beta3a with both standard and snapshot 
> > 20001219 versions of
> > > GoAhead.
> > 
> 
> 
> **********************************************************************
> 
> Check us out at http://www.syntegra.com
> 
> **********************************************************************



More information about the users mailing list