Re: Unable to load interpreter ???

[Sascha Schumann <sas schell de>]

On Mon, 30 Nov 1998, Stefan Boresch wrote:

> > 
> > I have seen this behaviour before.  If you have sustained heavy paging, your
> > system's virtual memory completely runs out and then is reclaimed back to
> > sensible amounts, in continuous oscillations.  You can watch this happen,
> > although tools like vmstat and top don't last for more than a couple of
> > hours under these extreme circumstances (I had a nice graph somewhere of vm
> > cycles on what should have been a steady-state overnight test).  Attempting
> > to run a shell when vm is near exhaustion produces the 'unable to load
> > interpreter' message.
> > 
> > The kernel was the patched 2.0.35 (may have also happened on 2.1.122, can't
> > remember).  The machines - LX and UX boards (don't have anything else to
> > test).
> > 
> 
> Hm, wait -- I accept this explanation of what is going on (my problems
> occur after heavy paging has been going on for a while...).  However,
> what can I do about it?  Are there patches/bugfixes whatever out there
> that fix this?  I am positive that I am never using more memory than
> the sum of RAM + swapspace (actually far from it).  The memory
> tester is run in such a manner as to consume just a bit more memory than
> there is RAM (mainly to make sure that every bit of RAM is actually
> stressed).  Our real applications use much less memory than there is
> RAM, but I guess because of the heavy disk I/O (virtual) memory gets 
> used a lot for buffers/caching etc.

You wonder that it cannot load anything without having memory? 

> But if I understand this right, than I am looking at a memory management
> bug, or not...?

It's not a bug. People seem to think every point they don't understand is
a MM bug.