[tech] morwong

[Grahame Bowland]

On Tue, Sep 19, 2000 at 12:10:39PM +0800, David Manchester wrote:
> > Sometime in the last hour or so, morwong started becoming less and
> > less responsive, and we were eventually forced to reboot it. CPU load
> > was pretty low, perhaps it was a deliberate or accidental forkbomb...
> > At the end, top took several minutes to start then immediately
> > segfaulted, "ps aux" did not complete and NIS lookups on other machines
> > were failing.
> 
> Any visible culprits yet?

Not really. It was so badly broken we couldn't run ps :) We might have looked 
at /proc but didn't have time.

> *groan* Why are we rehashing this?
> The UCC's getting to be a really frustrating environment to work with...
> 
> Tear down the machine room / don't tear down the machine room.
> Don't fix the monitors/air-con, buy a new one, without even getting quotes
> for repair.
> Throw away the nice Alphaserver and get a happy fun PC for all the 
> important things.

I totally agree with you. I wish we'd come up with a 'plan' and stick to it :)
We have money at the moment - we may as well work out what we want to do with 
it. Having four grand in the bank is stupid - we could just put it in a term 
deposit if we don't want to spend it now.

We should investigate fixing the air con. We really should. However, nobody has 
volunteered and I'm not going to - too much to do already.

And if we're going to anything with a PC, I think we should replace mermaid. 
We shouldn't buy a brand new Athlon (people have been talking about this) and 
justify it being unusable by most club members because its 'unstable'. How 
about we buy a nice coding box, run Debian/stable on it or even something 
not Linux (?BSD) and let people use it for coding or whatever makes them 
happy. We have mussel as the unstable box if people want to do fruity stuff.

> I've been away for four days... what has morwong been doing ?

Working fine. It's running reasonably well. Personally I think Nick was 
over-reacting. Sure, it's had some downtime. This has been true of all our 
boxes, because Cameron Hall power keeps tripping. It was on the same power 
segment as the main clubroom, which looses power more often than the main 
machine room. I fixed this.

Basically, it's been rock solid except for power failures and one incident. 
We think that was a fork bomb - probably accidental, and we can avoid that 
by putting process/memory quotas on. If we make them large they won't be 
overly restrictive.

Cheers,
Grahame

-- 
Grahame Bowland - http://gbowland.ucc.asn.au/