[SATLUG] lvm2 and raid 0
j at jvpappas.net
Sun Jan 6 09:03:36 CST 2008
On Jan 5, 2008 10:16 PM, Ernest De Leon <edeleonjr at gmail.com> wrote:
> I always use LVM2 for new server builds.
I wholly agree, and I also use it for many PC builds. The flexibility
inherent in the LVM construct is well worth any overhead (Complexity or
computational) incurred by it's use. Online `pvmoves` and snapshots are
priceless, especially in the case of servers. Throw in the capabilities of
advanced SAN arrays, and LVM is nearly a requirement.
Not only do you get the redundancy and performance increases,
I do not understand to what you are referring. AFAIK, LVM does not provide
any inherent data protection methodologies for redundancy. A RAID block
device managed by LVM could have redundancy, but not through the merits of
LVM. On the performance side, there are certain (Small OS files not being
one of them) where the extent size helps retrieve data more rapidly by
decreasing I/Os and increasing throughput, but on the whole, any
introduction of indirection rarely increases performance. We have used LVM
striping over multiple RAID LUNS to increase performance by spreading IOs
over multiple array controllers, but this can be risky (not to mention
I am interested in understanding your reference to LVM-enhanced redundancy
> you also get the ability to resize the LVs on the fly.
This is true, but only if the file system supports it. Nearly all
mainstream file systems support online growth, and nearly none of them
support online reduction. That is not to say that off line reduction is not
possible, as some of them have that capability through their native
toolsets. It is not (usually) as easy as an `lvextend` to manipulate
storage capacities, but needless to say, without LVM there would not even be
a discussion, as online disk partition manipulation is not viable.
Believe me, I agree whole-heartedly that LVM is an essential tool in the
sysadmin arsenal, but not a silver bullet.
More information about the SATLUG