[Beowulf] Big storage

[Greg Lindahl]

On Wed, Apr 16, 2008 at 02:50:14PM +0100, Tim Cutts wrote:

> Given that they state in that paper that they performed the same  
> analysis on a 2.4 kernel and the proportion of these errors was about  
> the same, I would be willing to bet the answer to that question is  
> "Not many, and there are probably plenty of new ones to counteract  
> those which *have* been fixed"

I don't think "they" have changed their philosophy -- if you don't
know what to do, you should at least log that you encountered that
weird case.

Those of you who use the PathScale compiler may remember the
occasional "Not yet implemented" error message -- it's a great way of
getting QA or a customer to provide example inputy to use to develop
working fixes for weird cases.

Filesystems are easy: you just inject errors, and you'll hit most of
the corner cases. Right now that means silently wrong answers or
crashes, with no clue as to where the problem is.

-- greg