For the past many months, one of my system has been rebooting
for no apparent reason.  Finally caught the panic message on
the console, unfortunately, no crash dump.

Anyways, digging the lists revealed many posts about similar
crashes, and i gather that this should be solvable by raising
NKMEMPAGES.  Something must have changed at some point as this
is a relatively new problem (for this system).

Anyhow, i have couple questions:

(1) how am i supposed to pick a better value for this?

(2) how can i monitor usage in a meaningful way?  vmstat -m ?
    Also, vmstat -m reports a total allocated of 37012K which
    is nowhere near vm.nkmempages * hw.pagesize so i'm confused.

this system is a sparc w/ 512MB of RAM, 128MB being used for /tmp
running 1.6 (compiled September 9th from a fairly fresh cvs
source).

vm.nkmempages = 1536
hw.pagesize = 4096


The other thing that really bugs me is that upon reboot, fsck
cleared 1124 files which i find rather astonishing considering
the light usage of this box of which i'm basically the only
user...  Most (all?) of them were of specific days rather than
spread out over time which makes me even more suspiscious of
something being wrong, but i can't really think of any way to
track down where they all came from or whether there are any
potential such files now.

any insight on all this would be appreciated.