Hi all,
Although I am a newbie to ADSM I would like to give some comments
to the discussion about ADSM and AFS.
First, I am pleased to see that our site is not the only one being in
front of the task bringing ADSM and AFS together. A broad interest on
customer side causes a real interest on vendor side.
On Thu, 19 Jan 1995, R. Alexander wrote:
>We plan to use an R24 w/1GB of memory for both interactive users
>and the ADSM server function. Can you see
>any problems with this arrangement?
Funny enough, right now we have started to install two R24 for this job.
To merge interactive load with file server functions seems to me dangerous.
Considering my experiences in the past I have not the impression
that the standard unix system can handle high interactive load properly and
do a good job for special services at the same time in the background.
This is why we decided to dedicate our R24 for file serving and nothing
else. Anyway, I should mention that I have no experience with AIX systems,
maybe this not true for AIX systems.
On Fri, 20 Jan 95, H. Pfleger wrote:
>The second problem (and this is the real reason why we don't use adsm to
>backup AFS volumes) is that adsm don't know, when it leaves the own cell.
>
>In our case, many User makes mountpoints from their home to points of
>interest all over AFS-world. This is a easy way to reach (user)important
>files. If you (our your user) starts backup, you will do this for many
>system:anyuser partitions all over the world.
So far we have no users with mountpoints to other cells in their directories.
From my point of view there is no reason to do so. Such things destroy the
clear hierarchical structure of the AFS file space.
Nethertheless this should be fixed by IBM and I suppose it can be done
without to much changes to the code.
On Fri, 20 Jan 95, Shyh-Wei Luan wrote:
>Current ADSM users are not defined across machines. One needs to
>do backup and restore from the same machine, otherwise, authorization rules
>need to be set up for cross-machine access.
This fits not very well to the slogan of AFS "no matter where you go ..."
Joe User must be able to retrieve his AFS file from any AFS client in his
cell or even in the world. But we have not to go so far away. Consider a
local workstation cluster tied together by NFS. Even in such a cluster it
is convenient to access archive or backup files from different workstations
in the cluster in an easy way. The concept 'one user - one node' is suitable
for PCs and single user workstations but not in a multi-user-environments
where often many users are forced to share a bunch of workstations.
This should be improved in the near future by IBM.
>I am writing a program that serves as an ADSM replacement to AFS "butc".
>This program works as a backend to the AFS backup command,
>and ships the AFS volume full/incremental dumps to an ADSM server.
>We are testing this program in our (Almaden) production cell.
I am very interested in such a program. We have the same problem as Hans
Pfleger - and many others, I guess. There is no mean to do an AFS backup for
disaster recovery to ADSM, i.e. to restore files in case of a disk crash.
With an interface to butc we can use the standard AFS backup utility for
this purpose. If possible I would like to get the code for beta testing.
We have an AFS source license, if this matters.
>I designed and implemented a prototype ADSM clent and server that
>uses the Kerberos authentication in AFS to allow cross-machine principals.
>I do not speak for the product division's planning in this direction, though.
This is the first time I heard that there are efforts in this direction
and I am happy to hear about it. From my point of view this is the only
proper way to go. Why to reinvent the wheel again and maintain a second
user database, password administration and so on, if the user can be
identified by a valid AFS token? Same is true for a pure kerberos
environment or a DCE cells.
I would be very interested to hear a statement from the product division
concerning this topic.
On Fri, 20 Jan 95, H. Pfleger wrote:
>All hints from IBM shows me, that this people (both, the Austrian support
>team and their coworkers in US) don't realy know, what AFS is and which
>tools a system administrator needs to do a real campus-wide backup in
>case of AFS.
My experiences with IBM are too new to be able to really jugde this statement.
At least some people seem to work on a solution as the reply of
Shyh-Wei Luan shows. And especially the idea to use kerberos for this job
is promising.
If at present many people at IBM have not the right view to AFS and ADSM,
it is our job as customers to show the importance of this topic. There are
a lot of backup and archive software product coming up, most of them not
having an interface to AFS. If IBM creates such an interface, ADSM will be
ahead one step in the Unix market.
Werner Baur
--
===========================================================================
===========================================================================
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
===========================================================================
|
