>I would like more information on how you are using adsm to
>backup your afs files. We are running afs here and I backup
>some of my afs files from my SUN. Because adsm for SUN does not
>support the backing up ACLs along with the files, this is only
>a partial solution. The best approach would be to be able to
>do the backup from the afs file servers rather than from
>individual clients.
The best approach would be to do it from the afs file servers and to bypass
the AFS cache manager, i.e. to go from the vice partition straightforward
into ADSM.
For the first item we have to wait until the dsmc.afs client will be
available, if ever, for non-AIX platforms. However, a client which is able
to backup AFS acls solves only part of the problem. Backing up AFS trees
by ADSM you loose volume header information (mountpoints and so on). So
this method is not suitable for recovery from disk crashes.
To address the second item you need a completely different approach. I am
currently evaluating a prototype of a modified butc which has been developed
by Shyh-Wei Luan, IBM Almaden. Beta tests has been promising so far, and I
hope the prototype will become a real product, so we can use it for our AFS
backup strategy.
I have to admit that until now we are not yet backing up our AFS data by
ADSM, but it is quite sure that I will switch to ADSM in some weeks.
I plan to combine the backup utilities:
- weekly or monthly backups by adsmcoord (the modified butc)
- daily incrementals by dsmc.afs
Of course this strategy has some advantages/disadvantages:
- backup of twice as much data as necessary
- dsmc.afs only available for AIX
+ fairly up-to-date data can be restored
+ restore is faster because whole AFS volume are restored from ADSM.
ADSM is very slow if a lot of small files are to be restored
Our ADSM-Server runs on a RS6000 and is also AFS server and client. So
we do the dsmc.afs right on this machine. All AFS servers are attached to
an FDDI backbone which increases performance of adsmcoord backups.
I am aware that this approach is kind of patchwork, but I cannot wait until
the perfect solution comes out, I doubt it will ever. Some keywords for a
more homgeneous approach would be space management, multi-resident AFS, DMI.
BTW: IBM Rochester, a fairly large AFS cell, has implemented a different
backup strategy: they feed ADSM with local files created by vos dump/restore.
Werner
--
===========================================================================
===========================================================================
Werner Baur
Leibniz-Rechenzentrum X.400: S=Baur;OU=lrz;P=lrz-muenchen;A=d400;C=de
Barer Str. 21 RFC822: Werner.Baur AT lrz-muenchen DOT de
D-80333 Muenchen Tel.: ++49-89-2105-8781
Germany Fax: ++49-89-2809460
===========================================================================
|
