On Tue, Jun 17, 2003 at 03:38:27PM +1000, matthew green wrote:
>    > i concur.  i basically stopped using _all_ sun4c class machines
>    > because of rare random lossage.
>    
>    My experience indicates that that's an overreaction.  I see no problem
>    with the more recent sun4c machines - the only ones I specifically
>    recall seeing trouble on are the 1, the 1+, the IPC, and either an SLC
>    or an ELC, I forget which: the ones with software-flush caches.  In
>    particular, the 2 and the IPX have never given me any trouble.
> 
> unfortunately, i had problems with both SS2's and IPXen.... i've
> had them basically forever.  never had a /sparc machine stay up
> more than 100 or so days until i had a sun4m machine.  they're

I've had very good luck with my IPC name servers (bind8), though
admittedly they run NetBSD 1.4.2 at the present.  My record is just a
few days less than a year (360+) days.  I can't really test more than
that because our school routinely shuts down the power to the building
once a year to do testing of the electrical system.  I ran an IPC with
1.5.1 for an entire semester (end of March to beginning of August) as
it temporarily took over as my on-campus web server (only light access
to apache).  It's on the basis of my experience with it that I'm going
ahead with the upgrade to 1.5.3.

I did have some crashes when I was running an early version of bind8,
but after updating to 8.3 it never happened again, so I figured bind
was the culprit.  The only problem I recall is trying to compile openssl.
Stangely, the same openssl gave me trouble (one core dump) in the recent
update.

I don't run X on these things; in fact I rip out all the video cards and
other sbus cards to keep them as open, and hopefully as cool, as possible.

I usually give them lots of swap, three times the physical memory of 12MB.
When I compile the kernel, I usually trim it down to just what I need, no
more.  Are there hardware/software factors that might be affecting the
stability (other than I don't push them to the limit)?

henry nelson