We are going through this exact same painful procedure. Let me get on my
soap box a bit and say to any TSM developers that this is a horrible
implementation. Unicode filenames have been support since 3.7, it was 4.1
that seems to have broken them for us and we did not know it for a long
time. Forcing new full backups on our 1/2 terrabyte server has been awful.
Now for the solution, your local error log (dsmerror.log) should tell you
every time one of these files is hit. To determine which client needs it
most, if you are running Wintel start with anywhere Macintosh files are
stored. After that if you are using roaming profiles I would migrate that
server because the profile information may have directory paths that are
not mapping correctly. From there on I would either check all my error
logs (not fun) or just convert the client systematically.
The way I worked around the file expiration issue was to install the new
client, but set up two backup schedules. The first would be the one you
have with no changes, and the second with a different node name would be
with the useunicodefilenames= yes and renamefilespaces=yes. After you run
this way for a month or so and have the majority of your file versions
(depending on your setup) you can rename both the nodes. The filespaces
carry the netbios name and therefore seamlessly map back to the old node
name. With the old node name renamed to some name like server1_archive,
you just never delete that node and you will have the versions in case you
need them. This procedure obviously works best with one or few file
servers.
Hope this helps.
Rob Schroeder
Famous Footwear
Reinhard Mersch
<mersch@UNI-MUENS To: ADSM-L AT VM.MARIST
DOT EDU
TER.DE> cc:
Sent by: "ADSM: Subject: Unicode conversion -
how do you handle it?
Dist Stor
Manager"
<[email protected]
.EDU>
08/06/2002 07:38
AM
Please respond to
"ADSM: Dist Stor
Manager"
Hello,
starting with version 4.2 the TSM client for Windows NT/2000/XP writes
filespaces in unicode format, if the server also is 4.2 or higher.
For new filespaces, this is ok. The problem comes with old filespaces,
which have been backed up with an old client or server version. They
have to be converted to unicode format, triggered by the AUTOFSRename
option on server and client side.
As long as this conversion is not done, some files are skipped during
backup (those having names containing characters from a code page
that is different from the current locale) and even better: if they
have previously been backed up with a pre-4.2 client, they are now
expired!
So, the conversion is URGEND! On the other hand, it is expensive,
because it renames the old filespace and creates a completely new
backup. For us this means that we cannot do all conversions at once,
we have to spread them over time. So now comes the question:
How do you determine, which filespaces need the conversion most?
How do you determine, which ones need it at all?
The skipped files do not show up in dsmsched.log (at least under
my 4.2.2 client running on Win 2000). Are there any other means,
perhaps Windows based, to find file name containig characters from
a different code page than the client's current one?
Has anybody gone through it at all?
Greetings,
Reinhard
--
Reinhard Mersch Westfaelische Wilhelms-Universitaet
Zentrum fuer Informationsverarbeitung - ehemals Universitaetsrechenzentrum
Roentgenstrasse 9-13, D-48149 Muenster, Germany Tel: +49(251)83-31583
E-Mail: mersch AT uni-muenster DOT de Fax:
+49(251)83-31653
|