[lkml]   [1998]   [Sep]   [16]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
    SubjectRe: Very bad swap bug -- 2.0, 2.1 at least
    On Wed, 16 Sep 1998, Rik van Riel wrote:

    > On Tue, 15 Sep 1998, Simon Kirby wrote:
    > > This swap bug that I mentioned a while back is still happening, and this
    > > time seems to be much worse than before. In this particular case it is
    > > happening on a medium-loaded web server running 2.0.35.
    > [SNIP]
    > > procs memory swap io system cpu
    > > r b w swpd free buff cache si so bi bo in cs us sy id
    > > 1 0 0 5376 18236 10200 52244 380 0 174 0 292 305 2 9 89
    > [SNIP 13 lines]
    > > 0 0 0 5376 16028 10200 53596 264 0 125 0 275 471 13 9 79
    > Strange... There is more free memory than there is stuff to
    > swap in. This does indeed suggest a bug, unless of course
    > you use perl for web forms (4 simultaneous forks which glob
    > the machine will have left some 16 megs.

    Some large perl script is probably what cleared the 16MB free. Still, it
    wasn't executing when I was running the vmstat (I was also running a

    > OTOH, if you have the values for freepages (/proc/sys/vm/freepages)
    > set to a rediculously high value, the swapping and paging might be
    > for real. Setting the values in this file to something like
    > "128 256 384" will leave you with some 1.5 MB of free memory
    > (which is a must when heavy forking is going on).

    "For real"? I'm very sure it is actually reading from disk -- I can hear
    the drive working every second, and if I run my "stail"
    (continously-updating cat) program on the /proc/scsi/aic7xxx/0 file, I can
    see the disk being read as well.

    [sroot@elmo:/root]# cat /proc/sys/vm/freepages
    280 420 560

    This is high, I know...should I try it lower?

    > > As you can see, the amount of stuff swapped out is staying the same, but
    > > it is constantly "swapping in" something. There is obviously a lot of
    > > memory available to have this data swapped back in and out of swap space,
    > > but it isn't happening.
    > The MM code in 2.0 was mostly introduced in the 1.2 (!)
    > days and is very well tested. Apart from the freepages
    > thingy or some hardware fault (very improbably, but then
    > again you're the only one to ever report this) I can't
    > think of many other errors.
    > Does the error also occur with a different kernel? Or
    > with a kernel compiled with a different compiler?

    All kernels compiled with It seems to be happening on 2.0.33 and
    2.0.55, and even on 2.1.120. I will try to write a C program that will
    reproduce it.

    > > >From the tests I did before, it seems what is being "swapped in" is the
    > > executable code of a program which has been previously swapped out. I had
    > Executable code is mmap()ed from the executable file, and is
    > not swapped in. Faulting pages from executables are shown with
    > the PageIn (pi) stats...

    "pi"? This does not appear to be in vmstats...Where is it?

    > > Issuing a "swapoff -a ; swapon -a" causes the program to go away, until
    > > something else forces some programs out to swap. (In this case, it seems
    > > to be some customer's run-away CGI script).
    > Running those is always a bad idea, but the system should be
    > resistant against that...
    > You might want to try 2.1.117 (running stably here) if
    > nobody has had very bad experiences with that version...
    > Or maybe the 2.0.36 prepatch (that one should be stable
    > too).

    I believe 2.1.117 was also having the same problem. We're using 2.1 on a
    web statistics compilation server due to how much faster it runs, and I
    also noticed the swap thing happening on it. I believe this was with
    2.1.117 when I noticed it for the first time.

    Thanks for your help,


    | Simon Kirby | Systems Administration |
    | | NetNation Communications |
    | | Tech: (604) 684-6892 |

    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to
    Please read the FAQ at

     \ /
      Last update: 2005-03-22 13:44    [W:0.035 / U:2.176 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site