Re: [PATCH] vm-fix-all_zones_ok (was Re: 2.6.3-mm3)

[Chris Wedgwood]

On Tue, Feb 24, 2004 at 01:22:22AM -0800, Andrew Morton wrote:

> Sigh.  There is absolutely nothing wrong with having a large slab
> cache.

Unless the slab contains useful data (ie. dentries that will actually
be accessed is some useful amount of time) then arguably those pages
are better used for the page-cache.

> And there is nothing necessarily right about having a small one.

Oh, I'm not claiming one is more right than the other.  The issue for
me is updatedb or even bk clone a few times will cause the slab to
swell with dentries and related contents that never seems to shrink
without forcing swapping.

> Sorry, but your comment is information-free.

I just notice I'm getting debugging output (call traces) for page
allocation failures which I've not seem before.  I was getting these
(swapper: page allocation failure. order:2, mode:0x20) after beating
things to see how bad things will get for Nick.

In this case slab is consuming so much low memory we're getting
allocation failures which was making networking choppy.


  --cw
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/