Hi,
I want to clone both UFS and ADVFS filesystems.
I don't want to use clonefset because:
- I can only make one clone per fileset at a time.
- a large number of modifications to the original fileset can lead to
"WARNING: advfs cannot copy-on-write data to a clone file"
and this would be very serious to our applications.
Does anyone out there use LSM snapshots for filesystem cloning (as opposed
to raw device cloning).
For UFS we can use:
===================
# volassist snapstart original_vol device
# volassist snapshot original_vol new_vol
# fsck -o /dev/rvol/new_vol
# mount -r /dev/vol/new_vol /snapshot
However, in volassist (8) it states:
snapshot
Some usage types will attempt to synchronize any in-memory data associ-
ated with the volume (such as unwritten file modifications) when the
snapshot operation is done. For ufs, the synchronization operation
consists of a call to sync(8), which will make the snapshot a better
image, but which may leave some inconsistencies between in-memory file
system data and the data residing on the backup image.
Does anyone have any practical experience of this? Is this operation
reliable or does it lead to corruptions?
ADVFS:
======
For ADVFS the problem is that the numeric ID of the filedomain on the
new volume is the same as that of the cloned volume and this confuses the
kernel. It's not possible to have both the cloned file domain and the
original file domain mounted simultaneously. Does anyone have a workaround
for this? How good is the snapshot operation at flushing in memory ADVFS
data. I've done some quick checks with /usr/field/msfsck and can't see any
problems. (although I've learnt to my cost that 'bug' and 'ADVFS' go together
like a horse and carriage).
Any war stories/advice/ideas would be appreciated.
Cheers
Keith McCabe
Banque Paribas Capital Markets
London W1
Received on Wed Feb 05 1997 - 14:46:18 NZDT