noflushd

[schnobs@babylon-kino.de]

Hello,

I've got a machine that ought to be capable to run it's tasks 
completely in RAM for long times (more detailsbelow).
free usually reports 30-40MB (out of 80) not even to be used 
for buffers.
Now, I thought there is a way to force the machine to fill it's 
buffers until either a) the disk's up anyway or b) it runs out of 
memory.
Furthermore, I thought noflushd would be taking care of this.

the command used:
noflushd -n15 /dev/hda

trying to check it with -vd, I get the following:
Error: no valid timeout for /dev/hda

Now I wonder what I'm doing wrong. Even more as the disk 
actually spins down frequently, but comes back to live every 
couple of minutes. Thats bad, is it?

More details and further ideas:
It is a more-or-less classical dial-in box.
It has a harddisk that from time to time is actually needed, but 
most of the time (approx 22hrs a day) its sole job is IP 
maquerading and running a nameserver (BIND9 as out of the 
box, caching-only).

ls -rt indicates that syslogd & al. are responsible for the 
frequent activity, at any rate the only files that have changed 
during the last couple of hours are to be found in /var/log.

Now, I don't want to live without logging, but consider to move 
the logfiles into a ramdisk and have logrotate copy them onto 
the disk. But this seems to be quite an endeavour, and if 
anyone can think of an easier way, please let me know.

cu,
Schnobs