[luau] Mililani's Sun server - (new) problems

[Gary Dunn]

On Wed, 29 Jan 2003 09:57:12 -1000
Warren Togami <warren at togami.com> wrote:

> The following is my message to the Department of Education's Cliff Goto.
> 
>     On Tuesday Rick Chavez and I spent another 8 hours at Mililani 
> dealing with these new problems.
> 
> Any ideas?
> 
> 
> =====================
> 
> Hi Cliff,
> 
> Over the weekend Mililani had power glitches, after which the Sun server
> would no longer boot.
> 
> * Some student turned the key into diagnostic mode, 
         ^^^^^^^

Careful. Students are often better at this than teachers. Try "some girl,"
"some Republican," "some jew." You get the idea? "Someone" would be a
better choice.


> 
> * Filesystem had hundreds of errors/corruptions.  We suspect some of 
> this was due to several unclean poweroffs and crashes, but a few times 
> we would cleanly reboot and fsck found even more errors.  Is this a sign
> of possible hardware failure?

Not necessarily. fsck cannot always end up with correct values after a lot
of fixups. If errors continue after three runs, I'd say you have a
hardware problem. Don't mount the drive until it's clean.

You DO have an UPS and an auto-shutdown on this box, right ;-)

> 
> * hme0 interface stopped working for seemlingly no reason.  We tried for
> 
> 5 hours straight until 9pm testing different settings with no success. 
> hme0 is the interface going to the school network/Internet.  hme1 and 
> hme2 both are working fine, serving two internal subnets of Sun Rays.
...
> 
> I'm hoping this is just some configuration issue that we're missing, but
> we tried for hours and couldn't get hme0 to work.  Could not even ping 
> the gateway.  Strangely 5 out of several thousand ping packets would 
> return, and the arp table sometimes showed the gateway and another host 
> on the hme0 interface, but nothing beyond this.

I have seen this with 100Base/TX when the NIC and the switch don't agree
on the protocol settings. The two issues are speeds, 10 vs. 100, and
half-duplex vs. full duplex. This is consistent with the power disruptions
you reported. You should reset the switch, either through
it's software interface or by cycling power. Then take down and bring up
the NIC ... does Sun use ifconfig?


-- 

          _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/
         _/                                     _/
        _/  Gary Dunn                          _/
       _/  Open Slate Project                 _/
      _/  http://openslate.sourceforge.net/  _/
     _/  http://www.aloha.com/~knowtree/    _/
    _/  Honolulu                           _/
   _/  registered Linux user #273809      _/
  _/                                     _/
 _/  This tagline is umop apisdn.       _/
_/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/