On Wed, Apr 19, 2023 at 06:00:53PM +0700, Max Nikulin wrote: > On 19/04/2023 13:34, to...@tuxteam.de wrote: > > On Wed, Apr 19, 2023 at 01:15:01PM +0700, Max Nikulin wrote: > > > Summary: Moving /tmp to tmpfs makes it useless > > > https://lists.debian.org/debian-devel/2012/06/msg00311.html > > > > > > That is linked from > > > https://wiki.debian.org/SSDOptimization/#Reduction_of_SSD_write_frequency_via_RAMDISK > > > > What I didn't like from the post is that it doesn't clearly > > state the downsides. > > Particular user may run applications requiring more space in /tmp than is > configured by default for RAM disk. Examples are image manipulation, > processing scientific data. The thread contains links to earlier > discussions. I admit that I have more free RAM + buffers than free space on > / as well, but it is a conscious choice.
Definitely -- but this will happen to you whatever device you have backing your /tmp. Arguably, you'll be better off if /tmp has an own device than when sharing one with /. > That is a link that I had. I do not have another one with a better balanced > opinion. My point that the suggestion may lead beginners to trouble, so they > should be warned. I don't mind opinions being unbalanced. They usually are (I know mine are). What miffed me was that they went over many paragraphs telling that it is bad whithout stating clearly /why/ they think it's bad. I had to pick up between the lines that they possibly mean you might run out of space. Well... yes? > Notice that debian wiki recommends tmp.mount systemd unit instead of > /etc/fstab line. Unsure if provides more flexibility or other advantages. No idea. No systemd over here. > P.S. Arbitrary time may be invested in optimization attempts. There are > interesting features like swap in RAM (compressed memory pages). Some people > get better performance for tasks like compiling chrome. Some can not achieve > it. Definitely,but we nerds gotta nerd :) Cheers -- t
signature.asc
Description: PGP signature
