[rescue] Last AIX 4.3.2 ML Media?

Fri Jan 20 08:35:43 CST 2006

On Fri, 2006-01-20 at 08:02, Eric Webb wrote:
> On Thursday 19 January 2006 12:29 pm, Patrick Finnegan wrote:
> > >
> > > Sounds like it ran out of memory and can't fork login.  That happens
> > > even on AIX 5.x when I try to copy a file larger than physical ram.
> > > It usually clears up once the operation hogging memory finishes.
> >
> > If that's the problem, you might try adding more swap space, and see if
> > that helps.
> 
> That's what I first thought as well, but now I doubt it.  The app on this 
> machine was moved over from an R50 with 1/3 less physical ram and has 
> adequate swap.  At some point, doesn't it start killing off processes just to 
> stay alive?  There could of course be a sharp upward spike right as this 
> happens that my monitoring doesn't see, but every metric I can find indicates 
> business as usual up until the problem.

I wonder if it isn't an arbitrarily-low kernel setting.  I am not an AIX guru, but aren't there kernel tunables like NPROC or MAXUSERS ?  Maybe you just need to increase the number of file handles or whatnot.

I came across this:
http://publib16.boulder.ibm.com/pseries/en_US/aixbman/prftungd/2365a82.htm

but it appears to be for 5.2, not sure.

http://publibn.boulder.ibm.com/doc_link/en_US/a_doc_lib/aixbman/prftungd/2365a82.htm

appears to have some details - I wonder if v_exempt_secs is set to a sane value?  Also maxpin maxpout maxrandwrt nokillroot and some of the IO stuff further down the page.

--Patrick