Salvo Tomaselli <tipos...@tiscali.it> writes:
>> No, tmpfs will be swapped out if you don't use a file for a while but
>> something else uses memory, including IO caching.
> unless too many things want to use memory, then tmpfs gives a great
> contribution in taking down the machine.
>
> As you pointed out yourself in another email, under memory pressure the
> kernel
> starts doing odd choices.
>
> So the point is: is it correct to enforce a default setting that will break
> many software that would otherwise work flawlessy, and that makes the machine
> less reliable but faster for certain kind of tasks?
You're still ignoring that a disk based filesystem puts up the same
memory pressure, even more so for the journal and metadata overhead.
As I pointed out the tipping point can be easily recreated by writing to
NFS. A slow USB device works too if you have one big enough. I haven't
seen the same behaviour with tmpfs but then again I don't put files much
much larger than my ram in /tmp.
MfG
Goswin
--
To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/877gvmdwv8.fsf@frosties.localnet
