ADSM-L

Re: user initiated restores on aix

1996-11-26 06:35:09
Subject: Re: user initiated restores on aix
From: Helmut Richter <Helmut.Richter AT LRZ-MUENCHEN DOT DE>
Date: Tue, 26 Nov 1996 12:35:09 +0100
On Mon, 25 Nov 1996, Bill Hunter wrote:

> Hello, I am trying to figure out how to give my aix v3.2.5 users the
> ability to restore THEIR files using either the adsm gui or linemode
> clients.  The Rel. 1 adsm user's guide and reference don't seem to
> have the information I need.

You have to define PASSWORDACCESS=GENERATE so that your users may use ADSM
without specifying either node name or password. In this case, they have
access to exactly their own files. If you have more than one node name,
you need more than one server name: for some reason, specifying a node
name is considered by ADSM as an attempt to log into ADSM as root user of
that node, and the password is required. More than one server name does,
however, not mean you have to run as many servers.  Rather, one server may
pass under more than one name.

If your users' data reside on a file server where the users have no
access, you have still to validate all your users there. This is a bug in
ADSM (which I have not reported so far and I wonder whether anybody else
has): The non-validated users' files are backed up from the server under
the users' UIDs instead of UNs (because the latter are unavailable to
ADSM) but a user is validated against ADSM via his UN and not UID. Thus
ADSM refuses the data ("this is user JoeUser but the data belongs to UID
23456", even if JoeUser has UID=23456).

If you are opening the service for a large institution such as a
university, be sure to split users as much as you can: not too many users
per node, not too many nodes per server, not too many servers per server
hardware. If this splitting is not necessary, you can temporarily revoke
it at any time by setting up the configuration in a way that in effect the
same server is always used. If, however, this splitting becomes necessary
while the service is already underway, you'll have a hard time to
introduce it afterwards.

These are the pitfalls we have had to circumvent (passwordaccess is easy,
only the documentation lacks precision; the other two persist here
because they are too expensive now). But in principle, things do work.

Best regards,

Helmut Richter

==============================================================
Dr. Helmut Richter                       Leibniz-Rechenzentrum
Tel:   +49-89-289-28785                  Barer Str. 21
Fax:   +49-89-2809460                    D-80333 Muenchen
Email: Helmut.Richter AT lrz-muenchen DOT de    Germany
==============================================================
<Prev in Thread] Current Thread [Next in Thread>