ADSM-L

Re: export import node metadata TSM database

2005-09-15 16:03:20
Subject: Re: export import node metadata TSM database
From: "Prather, Wanda" <Wanda.Prather AT JHUAPL DOT EDU>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Thu, 15 Sep 2005 16:02:58 -0400
Look at the server VOLHISTORY; export tapes are recorded there.

I believe you will find that if an EXPORT spills to a second tape, the
second tape is recorded as "vol seq #2" of the export, not as an
independent tape like storage pool tapes are.  I believe TSM has to
start reading at Vol seq #1.

I also know that most tapes of tape hardware supported by TSM don't have
the ability to record any positioning information, except for tape
marks, and I can't think of any case where TSM uses tape marks to
separate data on tapes.

"...if the full export is eg 2 TB and a restore is asked of a file that
is of course at the end of the latest tape in the export process"

Now I'm not trying to quibble with your terminology, but just to make
sure your customer understands:  you cannot RESTORE a file from an
export tape.

To get data back from an EXPORT tape, you have to IMPORT at a MINIMUM
the entire filespace you want.  That means reading and importing ALL the
data in the filespace, no matter how many tapes.  That rebuilds the DB
entries for that filespace.
If your IMPORT is eg 2 TB of SMALL files, it will take a LOOONG time and
have a big impact on your TSM DB.
And if an IMPORT fails (say because your DB isn't big enough!), you have
to START IT FROM THE BEGINNING, it is not restartable in the middle.

The most appropriate use of exports are
1) to move a node from one server to another, or
2) to take old data OUT of TSM and put it on a shelf, if it isn't likely
to be used again.  Because then you can delete the filespace from TSM,
which removes the metadata from the DB as well as getting the data out
of your tape library.

If your customer is just trying to create another backup copy at a point
in time, a BACKUPSET is more appropriate.
You can restore files from a backupset, without re-importing the data
into your DB.
(BACKUPSETS are also tracked in VOLHISTORY.)
But you still have to read the entire tape, and you can still put only 1
client per tape.
(There is no way to know what is ON a backupset, either. )

If the customer is actually only concerned about PARTICULAR data, say an
accounting data base, and not ALL the data (including operating system
files), the most appropriate thing is an ARCHIVE.  Archives are designed
to create point-in-time snapshots of data that will never change and not
expire until a defined expiration date, or until someone deletes them.
Put your archives in a separate storage pool from your backups, and you
can eject and send the tapes elsewhere.

If your customer wants ALL data monthly, and If you can afford to push
the data across the network again, the simplest solution may be to
create a second node-name for each client; put all those node-names in a
separate domain.  Schedule a selective (full) backup monthly to a new
management class in that domain that goes to a different disk pool that
migrates to a different tape storage pool.

Those tapes will "stack" multiple clients per tape, as you want, and all
the tapes can be ejected and sent wherever you want them to go.

The scheduling will be easy, the # of tapes will be minimized, restores
will be easy.
But it will, of course, take a lot of time to do full backups and will
impact your TSM DB size.
You could try doing IMAGE backups instead of filesystem backups; that
should be faster, but restores are much more problematic.
If your DB gets too big, you can spawn another instance of the TSM
server, just to service those nodes.

Pick your poison.  
Your customer is asking you to do the things that TSM was designed to
AVOID - i.e., create a lot of duplicate copies of the same data when the
data hasn't changed.  So it will be time-consuming and painful either
during the creation or during the restore!  

Wanda Prather
"I/O, I/O, It's all about I/O"  -(me)
   

    



-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU] On Behalf Of
Kurt Beyers
Sent: Thursday, September 15, 2005 3:12 PM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: Re: export import node metadata TSM database


Thanks for the answers Wanda.
 
Exporting all of the active backup data of all the servers in 1 command
isn't an issue of course, just one command and you are sure that you
don't miss anything.
 
But if the full export is eg 2 TB and a restore is asked of a file that
is of course at the end of the latest tape in the export process, would
this require then the full reading/searching of all the tapes? 
 
The restore procedure would then become a rather time consuming job if
there is no tape positiong info in the TSM database. Anybody with
hands-on-experience in such a setup? 
 
Or can you find out which tapes were used for which node in the export
messages and then just run the import on eg two tapes instead of
specifying all the used tapes? 
 
Else the export should be done on seperate tapes for each node resulting
in a high tape consumption and a lot of admin schedules of course. 
 
This would result that the export cannot be used in a practical way to
export full backups out of the expisting primary storage pool on a
regular basis (once a month due to the SLA agreements). 
 
I know of course the exoprt was not written for a such a task  and a
full backup is not in the TSM dictionary, but I'm just trying to be
creative to answer the customer wish list.
 
thanks,
Kurt

________________________________

Van: ADSM: Dist Stor Manager namens Prather, Wanda
Verzonden: do 15/09/2005 17:42
Aan: ADSM-L AT VM.MARIST DOT EDU
Onderwerp: Re: [ADSM-L] export import node metadata TSM database



-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU] On Behalf Of
Kurt Beyers
Sent: Thursday, September 15, 2005 9:54 AM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: export import node metadata TSM database


Hi,

Is there any metadata of the exports stored in the TSM database?

==> Yes, the volser is retained in the VOLHIST file. 

If I want to export eg two clients A and B to the same volume (eg a tape
vol001 that has enough space of course), can I specify twice the
command:

export node A filedata=backupactive devclass=<tape> VOL=vol001
export node B filedata=backupactive devclass=<tape> VOL=vol001 

==> I don't believe TSM will let you do that; it SHOULD tell you that
vol001 is already in use.


Or must I specify

export node A,B filedata=backupactive devclass=<tape> VOL=vol001

to have both nodes exported to the same tape?

==> Yes, that is way to get both on the same tape.


And what if want to run an import of node B only for a restore from the
tape that contains the export of node A too with the command:

import node B filedata=backupactive devclass=<tape> VOL=vol001

Will it go directly to the point on the tape where the export of B has
been written or will it read the entire tape untill TSM finds the
beginning of the export of server B?


==> If you say

IMPORT NODE Devclass=<tape> VOL=vol001
    TSM will import all nodes on the tape

If you say

IMPORT NODE B devclass=<tape> VOL=vol001
   TSM will import only node B

As far as I know, TSM doesn't save any tape positioning information that
would let it skip to the data from B; I think it will read the whole
tape.


Wanda Prather
"I/O, I/O, It's all about I/O"  -(me)

<Prev in Thread] Current Thread [Next in Thread>