Malcolm Herbert writes:
> 
> speaking of RAIDFrame, I recently came across the message from it to 
> the effect
> 
> help!  emergency full-stripe buffers depleted

Oh! Cool! :)
 
> ... not sure what that means ... anyone?

Many moons ago the memory allocation code in RAIDframe was changed so 
that it would have a chance to perform better in low-kernel-memory
conditions.  In particular, memory was allocated at startup for use
in cases where there was no more kernel memory available.  The above 
error indicates that your kernel became very short on memory, and
managed to use up all of the pre-allocated memory for the 
"full-stripe buffers".  (before the code was cleaned up, you probably 
just would have saw the system lock up :-} )

I think this is the first non-test case where this error message has 
been seen "in the wild".  Could you please send me (privately)
the output from 'raidctl -s' for the RAID set that error occured on? 
The 'magic constant' for the number of pre-allocated full-stripe 
buffers may need a bit of tuning, and so I'm hoping the output from 
'raidctl -s' might give me a bit more insight on how to tweak it.
Also: if you were doing anything else on that RAID set at the time 
(e.g. reconstruction or parity rewrite) I'd be interested to know 
of that as well.

Thanks.

Later...

Greg Oster