Original message:
> I have a strange problem with my DU 4.0B machine. Every once in a while,
my
> Apache, ircd, sshd and cron daemon processes die for no reason (that I can
> find). This usually happens when I'm not logged, which just adds to the
> pandemonium.
>
> The only clues that I've managed to piece together is that these daemons
are
> all standalone (ie: not spawned from inetd), and the following kern.log:
>
> Dec 23 10:31:02 nation1 vmunix: MACHINE CHECK type 0x660 Machine check
abort
<snip>
Conclusion:
I'm running out of swap space. Buy another hard disk and use a portion of it
for additional swap.
Dr. Tom Blinn <tpb_at_doctor.zk3.dec.com> wrote:
>This message means you had a hardware fault -- the system failed due to a
>hardware error.
Bryan Lavelle <Bryan.Lavelle_at_digital.com> wrote:
>Looks like you are having two problems, the one you show the panic for is a
>hardware problem, in no way a soft reboot. Your disappearing processes may
>be linked to running lazy mode swap. If you are running in lazy mode, as
>process starting up require swap space and it isn't available, other
process
>will be killed out of hand. No preference is given to any process to be
>kept alive. Check /etc/swapdefault and see if it is linked to your swap
>area, if not you are running in lazy mode.
Jim Belonis <belonis_at_dirac.phys.washington.edu> wrote:
>I would worry about your pagefile space allocation policy.
>There is 'eager swap allocation' and 'delayed swap allocation'
>If you have set up 'delayed swap allocation', then when a process
>actually uses swap space, it gets allocated. If there is not enough, then
>a random process gets killed to free swap space.
>if 'eager swap allocation' is enabled, then swap space is allocated
>when malloc() calls for it and if there is not enough, then the requesting
>process is told there is not enough space and typically would kill itself
>or run with less memory. A much more diagnosable situation.
Jim Belonis <belonis_at_dirac.phys.washington.edu> wrote:
>First consider whether this is indeed the problem. Watch the swap space
>when this happens if you are not sure it is possibly the problem.
>
>RAM will not fix swap. You would need more or bigger swap partitions
>(easiest would be to buy another disk and use a large chunk of it for
swap).
>Or you can change to the other space allocation policy.
>The other space allocation policy will only cause a more-diagnostic
>error message (the culprit program will die instead of a random program)
>and will not fix your fundamental problem.
>
>A secondary issue is, will your computer run too slow if you have small RAM
>and big swap ? If so, you indeed have to buy more RAM.
>(a typical rule of thumb is to make swap be 2 1/2 times RAM).
___________________________________________________________________________
ian wojtowicz <i_at_woj.com>
http://woj.com
Received on Sat Dec 26 1998 - 16:46:56 NZDT