On Wed, Nov 02, 2011 at 12:41:29PM +0100, Marek Podmaka wrote: > Wednesday, November 2, 2011, 11:18:50, James Hawtin wrote: > > Personally even with LVM I would still write a partition table the disk, > > this helps show the disk as being used to other system administrators, > > however there would be one partition on it of type 8e, and on that I > > would create a PV (physical volume), this PV can then be used to create > > How do you extend the PV then? For example extend the LUN on storage > or just resize a RAID1+0 set by adding 2 new disks... Resizing the > block device is no problem, resizing the PV also, but to resize the > PV, you need to resize the partition also - and if I remember well, > the kernel won't re-read the new partition table while it is used... That's just it -- normally you don't change the size of a PV. That's the reason for VGs, which can be expanded anytime by adding new PVs. To add capacity from a SAN, you could either: 1. create a new LUN, create a PV on it, and add it to a VG; or 2. expand the LUN, create a new PV in the added space, and add it to a VG. In case (2) you'd need to have partitioned the LUN before building the first PV on (a partition of) it, so that there can be multiple PVs on the LUN. There is weirdness in the SAN case because you have two LVM implementations layered up: one in the SAN fabric and another in the client host. Come to think of it: your hardware RAID probably has its own LVM implementation too, though it is probably a bare-bones one. -- Mark H. Wood, Lead System Programmer mwood IUPUI Edu Asking whether markets are efficient is like asking whether people are smart.
Description: PGP signature