kmem_map auto-sizing and size dependencies

[Adam McDougall]

On 1/22/2013 6:22 PM, Artem Belevich wrote:
> On Mon, Jan 21, 2013 at 1:06 PM, Pawel Jakub Dawidek <pjd at freebsd.org> wrote:
>> On Fri, Jan 18, 2013 at 08:26:04AM -0800, mdf at FreeBSD.org wrote:
>>>>   Should it be set to a larger initial value based on min(physical,KVM) space
>>>>   available?
>>> It needs to be smaller than the physical space, [...]
>> Or larger, as the address space can get fragmented and you might not be
>> able to allocate memory even if you have physical pages available.
> +1 for relaxing upper limit.
>
> I routinely patch all my systems that use ZFS to allow kmem_map size
> to be larger than physical memory. Otherwise on a system where most of
> RAM goes towards ZFS ARC I used to eventually run into dreaded
> kmem_map too small panic.
>
> --Artem
> _______________________________________________
Ever since related code has been patched to properly permit kmem=2x physmem
out of the box, I've been using that guideline for my systems and
I've not had an out of kmem panic in years.  I've previously ran into weird
temporary deadlocks with strong IO if I set the kmem too high, as if the ARC
caused something important to be overwritten; that is just my theory.  I 
agree
with the fragmentation principle here.  YMMV regarding the kmem size.