Messages in this thread | | | Date | Mon, 13 Oct 2003 09:28:51 +0400 | From | Hans Reiser <> | Subject | Re: ReiserFS causing kernel panic? |
| |
Ihar 'Philips' Filipau wrote:
> Hans Reiser wrote: > >> reiserfs is not warranted to work on corrupted hdds..... > > > Is there any kind of error statistics for hard drives? > > Geometry is known. > I suspect that structure of damages, caused by contact of plates > surface with head, can be classified.
> > > It may be possible to classify manufacturing glitches. I think HD > producers have this kind of classification/statistics - to improve > quality, keeping price low. > > Actually what I'm thinking of: some kind of design rules for file > systems, how to minimize crashing due to hdd glitches. > Let's say, if some of hdd regions are know to be more error prone - > desing fs to use those regions less. > If hdd damages used to have some specific structure - design file > system to keep renundant data in regions which are less likely to be > lost both at the same time. So renundancy would make sense. > > Is there any thing like this? > > Or file systems now do outlive hard drives?-) >
Block allocation policies affect performance a lot, and keeping them simple is important. I would however be interested in knowing what the distribution function for errors by geometry is. If it turned out that, say, errors were higher at the platter edges, I could make some format changes....
I think that for users it is best to think about how to mask drive errors in the device layer or the device using RAID and mirroring.
-- Hans
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |