The output from lvs isn't any simpler to parse really. This has the
same limitations as the existing implementation in regards to devices
moving so nothing is worsened with this patch.
I didn't see a reason to completely reimplement the whole method for
detecting lvm devices to get this bug that prevents booting in a very
common use case fixed.
--
Sent from my iPhone
On May 18, 2010, at 20:32, Alasdair G Kergon <[email protected]> wrote:
If this is LVM2, use pvs, vgs or lvs for queries - it should be
simpler!
See man pages for how to control output format. Remember PVs can move
on booting - it's the UUIDs found on the devices that matter, not
the device
names, so detection/waiting must be done during the actual boot.
Latest versions of LVM2 can also activate LVs while irrelevant PVs
in the VG
are missing, but it's risky to rely on that *if* booting after a
crash while
someone was changing any properties of the LV(s) concerned as the
change might
have been committed but not yet visible. (So only use that on static
configurations.)
Alasdair
--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]
