Re: webbased restore frontend?

On Wed, 2004-05-26 at 10:24 +0200, Miek Gieben wrote: 
> Hello,
> 
> In order to hack on something else than hdup, I've been thinking about
> a webbased restore "thingy". It should become something that an user
> can restore a file from a backup and get that file mailed to him/he,
> or that it can be downloaded from some server.
> 
> Is there any interest in such a thing?

If it came with a way to interactively select a set of files to be
restored, and then to kick off the restore process in the background, I
could definitely see it being useful.

The web interface per se is not the main thing, in my mind.  I'd first
work on allowing a set of files/directory to be restored in one shot
from the command line - currently you can only do a single directory or
a single file AFAIK -- and since it takes my system 3 hours to
decompress and search the backup files once, you really do want to be
able to extract everything you need in one run.  

Also it would be nice to be able to get a listing of files that are
available for restore from the command line, perhaps based on the
file-lists that tar writes out so it doesn't require
decompressing/decrypting the big files.

Once that logic is in place, adding on a "browse, select files, then
kick off background job" restore interface is not too hard - and it
would certainly be nice to have.  It could be done as a menu-driven
command-line interface like "restore -i"... perhaps you could first
browse through the file hierarchy, then select from the available
snapshot dates for each item that is to be restored.

Having a system that can look through indexes of what was backed up and
determine where snapshots of a particular file can be found could also
make basic restores much more efficient.  If the file you need happens
to be in the newest daily, and you know that ahead of time, then you
don't have to spend hours extracting the version from the monthly
archive only to overwrite the results later.

And then finally, once all that infrastructure is there on the command
line, you could also provide a web-based version of this -- which would
be nice for the poor IT guy who has a lot of careless users.

Many people, I suspect, will not want to take the security risk of
allowing apache (or any web server) any sort of access to the backup
files.  But one relatively safe way to do it would be to restore the
backup with original owners, groups, and permissions intact, and then
email the requester instructions to and copy the files they need from
the restore location.  If they didn't have access to them originally,
they still wouldn't have access to them after the restore.  I wouldn't
email the files themselves.

Bng
-- 
Boris Goldowsky <boris-FrUbXkNCsVf2fBVCVOL8/[email protected]>