On Fri, 22 Nov 2013 18:52:27 -0500, Paul B. Henson <henson@xxxxxxx> wrote:
Ah, that's true; using raid at the lvm layer allows you to selectively
choose redundancy as opposed to dropping the entire pv on top of an
mdraid.
Not sure I understood that passage...
But in any event I'm working on a script I'm calling "md-3par" which will
create a number of chunklets (aka LV) from a set of PV (contained
naturally in a single VG) onto which to assemble a MD array of chosen
geometry and redundancy level. Of course one can then additionally layer
LVM, bcache, or (b)tier as desired. LVM is just a replacement for the
archaic and inflexible partition table, IMO. Granted snapshots and
thin-provisioning are nice to have.
I am disappointed that LVM is trying to expand into MD's territory. It's
bad enough they had the LVM 'mirror' (IMO a sad effort, pitifully slow,
and given it's unability to handle missing mirror members very much not
useful) and an interleaved 'striping'. Simple concat/extension of course
makes sense but the One True Faith of *nix is one simple tool to do one
job and ONLY that job to the best possible. If they want to pretend to be
ZFS-lite then just do ZFS already but keep it out of LVM. </rant>
--
NSA - subscribe to our Timeline and you'll know even before they've hit
'ENTER'
Decency, Security and Liberty alike demand that government officials shall
be subjected to the same rules of conduct that are commands to the
citizen. In a government of laws, existence of the governmnet will be
imperilled if it fails to observe the law scrupulously.
-- Justice Louis D. Brandeis (Olmstead et. al. v. United States)
--
To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html