Hello,
I posted the same problem a fews months ago (using 3.0.2) and got no response. I hoped it would be corrected in 3.5, obviously it is not.
My way around this was to create the new VM from the template on local storage (or on shared NFS) and then run a script to migrate it to NFS datastore, because vmkfstools allows you to force the type of VMDK (-d thin). The script goes like this :
_________________________________________________
#!/bin/sh
#-- thin migrate from any datastore to an NFS datastore
#-- command line usage
if ; then
echo "usage : $0 vmname dstoresource dstoredest"
echo " vmname : name of VM"
echo " dstoresource : name of source datastore"
echo " dstoredest : name of destination datastore"
exit
fi
VMNAME=$1
DSTORESOURCE=$2
DSTOREDEST=$3
VMXPATH=`vmware-cmd -l | grep "/$.vmx"`
SOURCE=/vmfs/volumes/$DSTORESOURCE/$VMNAME
DESTINATION=/vmfs/volumes/$DSTOREDEST/$VMNAME
mkdir $DESTINATION
#-- unregister
vmware-cmd -s unregister $VMXPATH
#-- migrate from source
cd $SOURCE
for file in *
do
if (echo $file | grep "flat.vmdk"); then
#-- flat vmdk file, no action
:
elif (echo $file | grep "vmdk"); then
#-- vmdk file : thin provisionning
vmkfstools -i $file -d thin $DESTINATION/$file
else
#-- other file : copy
cp -p $file $DESTINATION
fi
done
#-- register new location
vmware-cmd -s register $DESTINATION/$.vmx
____________________________________________________
We get great space savings using this method, as we have about 30 VMs per NFS datastore. As you well know, it would be a shame not to take advantage thin partionning on NFS. The same goes for cloning : first clone anywhere, which will result in thick format event on NFS, and the use the script to thin migrate it.
--
Pierre
