Re: kswapd taking 100% CPU - only there's no swap?

["Stephen J. Smoogen" <smooge gmail com>]

On 3/7/06, Don MacAskill <don smugmug com> wrote:
>
>
> Arjan van de Ven wrote:
> > On Mon, 2006-03-06 at 16:27 -0800, Don MacAskill wrote:
> >> We have an 8-way RHEL4 AS system (4 dual-core Opterons) with no swap
> >> (because swapping is still broken for DB loads.  It swaps the DB out in
> >> preference of disk cache).
> >
> > zero swap is a really bad idea. At least make a 10Mb or so swap.
> >
> >
>
> I hear you, and will try it, but can you explain why zero swap is a bad
> idea?
>

Could be lots of reasons.. a lot more info would be needed to answer it:

What is your CPU
  Is it 32 bit or 64 bit
  How many do you have
  How much CPU cache do they have
  How does this particular architecture deal with more than X ram (NUMA, etc)
  How large are the processes that you are dealing with (1 gig, 3 gig, larger?)
  How large does this particular architecture deal with X amount of
ram allocated to a process?


> I don't want any disk activity.  I have ample RAM (32GB, of which I
> leave >3GB free at all times).
>
> I have plenty of diskless boxes which obviously have no swap.  Are you
> saying that I can't/shouldn't run diskless either?
>

I think you are putting words in Arjan's mouth here.. Arjan doesnt
work for Red Hat and is trying to be helpful for you.

> Why on earth would no swap be such a bad idea if I'm no-where near
> starved for RAM?
>

Well if a process gets larger than what certain slab tables.. I could
see it trying to be swapped off to disk because the process can't see
all of the process at one time.

> Don
>
> --
> nahant-list mailing list
> nahant-list redhat com
> https://www.redhat.com/mailman/listinfo/nahant-list
>


--
Stephen J Smoogen.
CSIRT/Linux System Administrator