David Härdeman wrote:
On Thu, Aug 24, 2006 at 09:27:49PM +0200, Ronny Aasen wrote:
Index: debian/partman-auto-lvm.templates
===================================================================
--- debian/partman-auto-lvm.templates (revision 40192)
+++ debian/partman-auto-lvm.templates (working copy)
@@ -37,6 +37,15 @@
You can choose to ignore this warning, but that may result in a
failure to
reboot the system after the installation is completed.
+Template: partman-auto-lvm/purge_lvm_from_device
+Type: boolean
+Default: false
+_Description: Remove all Logical Volume Configuration?
+ Before creating any Logical Volume Groups, we need to remove all
exsisting
^^^^^^^^^
existing
+ Logical Volumes and Volume Groups from the disk we are partitioning.
In general, I think phrases such as "we" should be avoided, but it
would be better if Christian/Kamion/joeyh could review the template.
+ .
+ WARNING This will erase all data on LVM Partitions
This will erase all data on the LVM volumes and on the disk?
+
Template: partman-auto-lvm/pv_on_device
Type: error
_Description: Existing physical volume on the selected device
Index: autopartition-lvm
===================================================================
--- autopartition-lvm (revision 40192)
+++ autopartition-lvm (working copy)
@@ -26,10 +26,52 @@
log-output -t update-dev update-dev
fi
+
# Check if the device already contains any physical volumes
realdev=$(mapdevfs "$(cat $dev/device)")
if pv_on_device "$realdev"; then
- bail_out pv_on_device
+ # Ask for mermission to erase LVM volumes + db_set
partman-auto-lvm/purge_lvm_from_device "false"
I think db_set will override preseedings (as Geert said)...
+ db_input critical partman-auto-lvm/purge_lvm_from_device
+ db_go
should be db_go || exit 1?
if you say so :)
+ db_get partman-auto-lvm/purge_lvm_from_device
+ if [ "$RET" = "true" ] ;then +
+ targetvg=
+ #what volume groups is on any of the the disk partitions.
+ all_volume_groups=$(vg_list)
+ #we only care about vg's on the selected disk
+ for vg in $all_volume_groups
+ do
+ if [ "$(vg_list_pvs "$vg" | grep -c "$realdev")" != "0"
] ; then
this looks fishy...will it actually return zero? Shouldn't this be:
if $(vg_list_pvs "$vg" | grep -q "$realdev"); then
grep -c returns a number of lines matching. 0 if none do. but -q will
work as well and probably be a cleaner way to do it
+ targetvg=${targetvg}" $vg"
weird quotation marks.....targetvg="$targetvg $vg" would be expected here
+ fi
+ done
+ for vg in $targetvg
+ do
+ #make sure the volume groups on the target disk
don't span any other disks.
very good idea
+ if [ "$(lvm_get_info vgs pv_count "$vg")" != "1" ] ; then
+ log-output -t partman-auto-lvs vgs
+ bail_out pv_on_device
+ fi
+ done
+
+ #it should now be safe to remove the vg's on the target disks
+
+ #remove lv's from the target vg's.
+ for vg in $targetvg
+ do
+ for lv in $(vg_list_lvs $vg)
+ do
+ #remove the logical volumes on the volume group
+ lv_delete $vg $lv
+ done
+ #remove the volume group
+ vg_delete $vg
+ done
+ else
+ bail_out pv_on_device
+ fi
in addition to lv_delete and vg_delete, a pv_delete should also be
done. It needs to be added to lvm-tools.sh, and should call pvremove,
I've just added it myself :)
fi
choose_recipe "$free_size" lvm || exit $?
After having looked at this, I'm wondering whether this shouldn't be
added to wipe_disk() in partman-auto/auto-shared.sh instead. If it
was, that would mean that all methods would be able to warn that they
are about to remove LVM VG/LV/PV's before actually doing so.
Of course, it would need to include a check whether the lvm-tools.sh
script is present before doing so, but I think it would be a good idea
in general.
I've attached a patch which details what this could look like...
Thanks for your hard work on this and sorry that I've jumped in with
my comments so late :)
Regards,
David
As long as it happens, it's good :)
thanks for reviewing and fixing, hope removal of lvm volume will be
included in next partman :)
Regards
Ronny
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
