In message <199911152113.WAA20014@mail.wins.uva.nl>, Frank van der Linden write
s:
>I've brought Kirk McKusick's trickle sync + FFS soft update code into
>the main tree. It has had some moderate to heavy testing over the
>last month. I'm currently using it on the /usr filesystem and the
>filesystem containing all my sources on my main NetBSD development
>machine, and did not have a problem since I fixed the last known
>bug (2.5 weeks ago).

On BSD/OS systems, there have been occasional chicken-and-egg problems with
needing to allocate memory to start doing anything that would free up memory
under *really* heavy load.  The answer support gives is to increase KMEMSIZE,
and indeed, this works beautifully - I'm wondering if the same advice is
likely to apply?

-s