23.07.2011 07:31, maurice wrote: > On 7/21/2011 5:18 AM, Stan Hoeppner wrote: >> ..Ok. Pol, keep in mind that all drives must be identical size when >> creating an array on raw disks. > That always makes me nervous. And this is not really correct statement. If you want to be sure, use --size=xxx argument when creating the array, and specify a size smaller than size of your drives. This way, you can avoid possible problems with drive size fluctuations and different size of a replacement drives from another manufacturer. Also, the "fluctuations" - i mean, several drives in the same batch may have slightly different size - will be masked out by mdadm itself, due to rounding to chunk size. Just use a few gigs less than your drive size is. Maybe the resulting size will be a nice power-of-two value, as well. > I have seen the case before where a drive has been "kicked out" and when > you go to add it back, it is refused as "too small" > Grown defects have shrunk the device. This CAN NOT HAPPEN. Drive never changes its size. Newly discovered defects gets remapped to _reserved_ space, not to a space taken from other data. If the drive runs out of reserved space (but you should replace it WAY before that), it will stop remapping sectors. > I find it is safer to make one big partition on each device, and make > this a bit smaller than the device. That works too, ofcourse. This way, it might be a bit easier to identify your drives too - for example, you can use GPT partition table there with proper labels for your partition. /mjt -- To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
