> >  [...]
> > > > The number of seeks should be closer to 1000 or 1500 per second, I
> > > > believe.
> > > >
> > > >
> > > > Again, it *flies* in single-user mode.
> > > >
> > >
>
> It would probably be a good idea to go into single-user mode and run, by hand,
> each of the rc.d scripts in the order that rcorder puts them in. After each

I suppose that that's the way to narrow it down.  Is there an easy way to
find out what scripts are being run in which order?

(Hm.  If memory serves, the last time that I had time to fiddle with
this---early January or late December---dd raw reads from the disk, even
in single-user mode, were very poor.  But it still is enviously (relative
to my current situation) fast if I mount from single-user before doing
anything else.)


> one runs you should check disk performance. This would give you a better idea
> of which part of the multiuser boot process is slowing everything down.
>
> I assume that during the time when you're getting dead slow disk access there's
> nothing else using the disk? (Sorry I missed the beginning of the thread)
> Something like syslogd might have gone insane and is writing constant logs,
> or somesuch.

I didn't say that; I probably should have.

Disk activity is 0 (LED is off) when I'm not doing anything else.

CPU activity can be very low (2%), so it's not that the CPU is too busy to
keep the disk occupied, either.

XOSView reports near-0 disk activity when doing things like starting
Navigator (which takes something like 45 seconds to 2 minutes to open it's
window---during which time, the LED on the drive is *constantly* on.)


  ``I probably don't know what I'm talking about.'' --rauch@math.rice.edu