>>That's interesting. Does it do the backup from the AFS server, or
>>is the backup server an AFS client, which scans down the AFS tree?
>>I think it makes a fairly large difference in terms of network efficiency.
>>xev
I don't know all the details, but it sounds like with ADSM, backups
are done based on the client system. The AFS files are just viewed
as more files owned by a particular user and get backed up with
their other data from the local system. Since ADSM wasn't written
with the AFS model in mind, it tries to extend the model of client
system backups to include AFS. Here is the question/answer data I
picked up from the conference. This is about the extent of what
I can tell you.
Thanks, Dawn
------------------------ from ADSM/AFS info sheet --------------------
Q. What call level of AFS does ADSM support?
A. ADSM assumes AFS users are already logged on. Otherwise,
ADSM will not be able to see the user's AFS directories
or files and cannot back up these AFS files.
ADSM treats AFS files almost like normal UNIX files, except
that ADSM saves ACL (access control list) information.
When ADSM backs up AFS files, ADSM saves ACL info by
using AFS system calls.
Q. Can root user on an AFS server machine backup all AFS client
files and let AFS clients restore their own files?
A. Yes and no -- ADSM relies on UNIX relationships and AFS
relationships to determine who has access to files.
How ADSM works in UNIX
======================
a. An ADSM client is a physical machine. Usually, the ADSM
client name is the physical machine name.
b. On a physical machine, more than one UNIX user can reside.
c. A root user can back up any files that reside on the physical
machine, regardless of who the owner of the file is.
d. When a file is backed up, ADSM saves the owner name of the
file, allowing the owner to restore that file. This means
that root user does not have to restore files for other users
on the same machine.
e. In addition, any user can give access privilege to other
users on other UNIX machines. This allows users on other
UNIX machines to restore files that were originally backed
up by another user.
How ADSM works with AFS
=======================
a. In order to use ADSM, a user must log onto AFS. We recommend
that the AFS user id be the same as the UNIX user id.
b. After a user logs on to AFS, all files belonging to that AFS
user id are now recognixed by ADSM (even if those files reside
on a different physical system).
ADSM stores these files under the following naming structure:
/afs/cell/usr/uid/filename
c. When a user restores an AFS file, ADSM attempts to restore
the file to the same naming structure. Usually this means
the file will be restored to the same machine from which it
was backup up. However, ADSM does not know what machine the
file is physically restored to.
