tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

re: very strange and severe raidframe performance issue

> i have a system where raidframe accesses are severely restricted
> and present a 2-20x performance loss.

to followup on this:

the problem has disappeared from normal raidframe accesses even with
the same kernels that were previously not working great.

i've since had the disk i was using both the underlying filesystem,
or when the mirror was in single sided mode fail [*], and i have been
able to confirm that the "slow parity rebuild" problem appears to
also be helped by a higher HZ setting.  with HZ=100 resyncing to the
replacement disk tops out at 6.4MB/sec (100*64KB).  with HZ=1024 it
is more like 62MB/sec (just under 1000*64KB).  i haven't actually
instrumented this to confirmm it but it seems likely.

[*] since the accesses to this disk were always fast from the
underlying disk except for a special case mlelstv has concocted,
and that the general slowness disappeared before the disk failed,
i am confident that the failure isn't related.


Home | Main Index | Thread Index | Old Index