Re: [gentoo-user] /var/tmp on tmpfs

[Rich Freeman <rich0@gentoo.org>]

On Thu, Feb 8, 2018 at 2:17 PM, Dale <rdalek1967@gmail.com> wrote:
> As someone else pointed out, if you
> start using swap, that generally defeats the purpose of tmpfs.
>

I'll just add one thing to this, which I've probably already said ages ago:

In an ideal world swap would STILL be better than building on disk,
because it gives the kernel fewer constraints around what gets written
to disk.

Anything written to disk MUST end up on the disk within the dirty
writeback time limit.  Anything written to tmpfs doesn't ever have to
end up on disk, and if it is swapped the kernel need not do it in any
particular timeframe.  Also, the swapfile doesn't need the same kinds
of integrity features as a filesystem, which probably lowers the cost
of writes somewhat (if nothing else after a reboot there is no need to
run tmpreaper on it).

So, swapping SHOULD still be better than building on disk, because any
object file that doesn't end up being swapped is a saved disk IO, and
the stuff that does get swapped will hopefully get written at a more
opportune time vs forcing the kernel to stop what is doing after 30s
(by default) to make sure that something gets written no matter what
(if it wasn't deleted before then).

That's all in an ideal world.  In practice I've never found the kernel
swapping algorithms to be the best in the world, and I've seen a lot
of situations where it hurts.  I run without a swapfile for this
reason.  It pains me to do it because I can think of a bunch of
reasons why this shouldn't help, and yet for whatever reason it does.

-- 
Rich