On Aug 13, 2010, at 3:50 AM, Rudi Ahlers <Rudi at SoftDux.com> wrote: > Hi all, > > Can anyone please tell me, from experience, if it's possible to > upgrade the hard drives in a RAID 10 system from 250GB HDD's to 500GB > / 750GB HDD's, while the server is running? The server runs CentOS 5.5 > x64. > > Our hardware vendors simply always say no, so we never actually tried > doing this, and when a client needed an upgrade like this we had to > build a new server with the same spec and larger hard drives and then > migrate the data across, which can sometimes take a few days to > complete (due to DNS & MAC address caches). > > So, I'm wondering, is it possible (and safe) to upgrade the existing > drives to larger drives in a running server? > > > I know we're done this with Linux's software RAID many times with > great success. But I don't want to take a risk if it's not possible. > The amount of effort and upfront costs for a new server makes hardware > RAID far more expensive than it's reliability, which Linux's software > RAID offers equally so I'm beginning to think we should just move all > the servers to software RAID in the future as well. If you create a bunch of RAID1s on the HW controller and use LVM to stripe them into a single VG, you can then pvmove LVs off a RAID1, take it out of the VG, replace it with a larger RAID1, re-add it to the VG, pvmove the data off the next one and so on. This will require you to either keep 50% of the space free for moves, or free slots open in the disk enclosure. You will also have to manually resilver the LVs afterward to have them striped properly (create new LVs and copy/move the data over). -Ross