On May 3, 2004, at 4:04 AM, Brad Knowles wrote: > At 10:00 AM +0200 2004/05/03, Erik Trulsson wrote: > >> Try that calculation for a machine with 32MB RAM. You probably need a >> minimum suggested amount of memory as well for that. > > Okay, then memory * 1.25, with a minimum of memory + 256MB, and a > maximum of memory + 1GB. Your suggested maximum seems like it would have little effect, percentagewise. What's the real difference to someone with an 8GB server whether their /var is _by default_ 9GB or 10GB? Presumably, however, they will notice that their suggested minimum /var size is larger than the suggested minimums of all the other partitions combined, and 9GB up from 256MB. If we increase default /var sizes, we should definitely tell the user why somewhere in the install process, so that they can change it back if necessary. Not everyone will want to sacrifice this amount of memory for crashdumps, for the same reason that people turn off INVARIANTS: the possibility of being more useful in a panic just isn't worth the cost, be it a gig of HD space or half the CPU being eaten by the system. There's also a privacy issue, as crashdumps may contain passwords and the like. Mike Hamburg >> A base value + memory size sounds better. > > As you proved, you need a lower bound of usefulness. But you also > need an upper bound, and when it comes to things like logfiles and > other things you might want to store in /var, the larger the memory on > the system the larger the logfiles, etc... are likely to be. > > Therefore, a multiplier based on memory is more likely to be useful > than simple addition. > > -- > Brad Knowles, <brad.knowles_at_skynet.be> > > "They that can give up essential liberty to obtain a little temporary > safety deserve neither liberty nor safety." > -Benjamin Franklin, Historical Review of Pennsylvania. > > SAGE member since 1995. See <http://www.sage.org/> for more info.Received on Mon May 03 2004 - 05:03:39 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:52 UTC