On Mon, 2005-09-26 at 23:57 +0100, James Wilkinson wrote: > Michael A. Peters wrote: > > Nothing in /tmp should be expected to survive a reboot. > > In fact, some people like to run /tmp as a tempfs (temporary file > > system) - only problem is, a tempfs can get full if you aren't careful. > > So can a regular /tmp. > > Modern tmpfs uses both memory and swap-space, depending on "memory > pressure" (whether the kernel thinks it has a better use for real memory > than using it for a particular tmpfs page). It has a > sysadmin-configurable maximum size: "the default is half of your > physical RAM without swap", but you can dynamically change that. Yeah - for the desktop user though, if you are doing a lot of downloading it can end up being a real pain if your download agent uses /tmp and suddenly you lose the download because your tempfs ran out of space. For desktops, I think it really is best for /tmp to be part of your root logical volume group. That being said - when I did sysadmin work, both /tmp and /var/run were tempfs (/var/run was tempfs for faster access to lock/pid files - and given a reboot after kernel update, you would never have stale pids blocking a service. How much of a difference it really made, I haven't a clue. Just what I was taught by dad - but it may not matter these days). Running /var/run on tempfs is a bit tricky in RH though, after it mounts you have to create directory structure that certain init scripts will expect to be there.
