Hi Keith, On Fri, Apr 26, 2013 at 08:18:40PM -0700, Keith Keller wrote: > Hi all, > > For other reasons I moved one of my md servers to kernel 3.8, and on a > whim I decided to test --replace on a RAID6 array. For this test I used > mdadm from Neil's git repository, but from a read of the md kernel > documentation you could probably do this just from manipulating > /sys/block/mdN/md/ appropriately if you don't or can't get the latest > mdadm. > > It went very smoothly in my test. I marked one drive with --replace, > and the array started rebuilding with the new drive, without any IO from > the other members of the array. It took much less time than a full > rebuild, and when it was done, the old drive was marked faulty, as > expected. > > This is really an amazing feature! I can see at least two use cases for > it: > > ==you suspect a drive is failing but it hasn't been kicked yet > ==you want to replace all the drives with larger drives in order to grow > the array > > Are there any other use cases? In the scenario where the drive marked my further idea would be to "rotate" a spare, so that the different HDDs will have, in the end, different power on hours. Of course, the spare must be idle while not in use. bye, pg > as wanting replacement fails completely, does md fall back to a full > rebuild with the new drive? Can it pick up in the middle or does it > have to start over? > > Thanks to Neil and the other md developers for such a great feature. > > --keith > > -- > kkeller@xxxxxxxxxxxxxxxxxxxxxxxxxx > > > -- > 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 -- piergiorgio -- 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
