Veritas-bu

[Veritas-bu] overwrite tape

2001-05-23 11:56:19
Subject: [Veritas-bu] overwrite tape
From: scott.kendall AT abbott DOT com (scott.kendall AT abbott DOT com)
Date: Wed, 23 May 2001 10:56:19 -0500
Yes, you are correct.  Assuming you have left the defaults to not mix
retention levels on the same tape (which should not be changed), the media
server will check to see if it has a tape available that meets the requested
retention level.  If so, it will use it until it fills up.  If not, it will
get a new tape assigned to it and start using that.  Each time it writes an
image to the tape, it adjusts the expiration date for the tape.

For example, a media server needs to do a backup with retention level of 5
(which, let's say is set to 1 year) from the NetBackup pool.  It checks to see
if it has any available media (not full, frozen, suspended, etc. ) already
assigned to it with a retention level of 5 in that pool.  Let's say it does
not find one.  It then requests a media ID from the VolDB host and is assigned
one if any are available in that pool.  If not and you are using a scratch
pool, one will be pulled from there.  Let's say we get media ABC123 assigned.
The first image is written to ABC123 on Jan 5, 2001.  The expiration of the
image is set to Jan 5, 2002 and the date the tape is set to be unassigned is
set to Jan 5, 2002 (not to be confused with the expiration date of the tape in
Media Manager, which tells you how long the tape has been in the system and
should no longer be considered reliable).  One week later, on Jan 12, 2001,
the same media server needs to do another backup using the same retention
level and the same pool.  ABC123 is still available (let's say it hasn't been
used since the last backup on the 5th), so the media server uses it.  This
backup is the second image on the tape and gets an expiration date of Jan 12,
2002.  The date the tape will be unassigned is now changed from Jan 5, 2002 to
Jan 12, 2002.

A year later on Jan 5, 2002, the first image expires and is removed from the
catalog.  It is still on tape and has not been overwritten yet so it could be
imported back into the catalog if necessary.  A week later, on Jan 12, 2002,
the second image expires and is removed from the catalog.  Also, the tape has
met it's expiration date in NetBackup (not the expiration date in Media
Manager) so the tape becomes unassigned from the media server.  The tape is
now available to be re-used using the whole process above, possibly with the
same media server, possibly with another.  At this point the tape is used from
the beginning and the old images can no longer be imported.

This is why it's not a good idea to change the default and allow media to hold
mixed retentions.  Several images with short retentions could be written to
the tape, but one image with a long retention could cause the tape to not be
reused for a long time.

This is in one of the manuals... either the NetBackup or Media Manager Sys
Admin Guide or the Troubleshooting Guide.

No offense, but this isn't Backup Exec/ARCServe/NT Backup and tapes aren't
labeled ServerA_Mon, ServerA_Tue, etc.  The whole purpose of using barcodes in
a library is to allow the software to track the media, saving the admin time.

If you only want one backup to go to the tape and then want it to sit there
until it expires, you can do that by scripting something that "suspends" the
tape after it is written to.  A suspended tape can be used for restores, but
can not be written to.  Unlike a frozen tape, a suspended tape will unassign
itself and become available after it expires without any admin intervention.

Now, if there is a requirement to ensure tapes are overwritten or destroyed
after x days, there have been many posts of how people are handling this.
Most include re-writing the label or degaussing the tape.  If you are trying
to do this, it is important to remember that suspending the tape as mentioned
above, will cause it to unassign when it expires, but does not mean that it
will immediately be overwritten.  It may sit in the library for weeks before a
media server requests a new tape and it is chosen to fulfill the request.
During this period, images could be imported back into the catalog and be used
for restores.  One thing that could be done is to create a pool for the
servers that have this requirement and to suspend the tapes in this pool after
each day's backups.  The pool will grab volumes from the scratch pool as long
as it needs.  If you had two media servers with a 21 day requirement, the pool
would eventually have 42 tapes in it (assuming each day did not use more than
one tape).  Once it hit this number, 2 tapes will expire and become unassigned
each day (suspended tapes don't need admin intervention for this and
unassigned tapes remain in the pool that they are in).  That night's backup
would grab these instead of getting new media from the scratch pool, thus
overwriting the media on the 22nd day (assuming it is in the library... you
need a library with a lot of slots to do this).  If you add a new media server
to this scheme, the number of volumes would increase, grabbing them from the
scratch pool, until the same thing started again with 63 tapes.  If less tapes
start being used because the data being backed up got much smaller or the
number of media servers was reduced, you would have a lot of available tapes
in the special pool and the tapes would not always get overwritten as they
became unassigned, however, all the tapes would remain in the special pool so
you could report on this with a script and take care of the extra tapes as
needed, reduce the number of tapes in the pool, and start automatically
overwriting tapes as they are unassigned again.


- Scott



                                                                                
                                                   
                    "Marcel van den Berg                                        
                                                   
                    (Devote)"                            To:     "Jewell, Jason 
T" <jj151001 AT exchange.daytonoh.ncr DOT com>,           
                    <marcelb AT interxion DOT com>              <veritas-bu AT 
mailman.eng.auburn DOT edu>                                       
                    Sent by:                             cc:                    
                                                   
                    veritas-bu-admin AT mailman DOT eng.        Subject:     
RE: [Veritas-bu] overwrite tape                              
                    auburn.edu                                                  
                                                   
                                                                                
                                                   
                                                                                
                                                   
                    05/23/2001 07:33 AM                                         
                                                   
                                                                                
                                                   
                                                                                
                                                   




Hi Jason,

I followed this procedure when I noticed the tape was full. As we have
just one tape available for each daily backup, the tape will be full
after one job (so each day the tape is full). To repeat the procedure of
executing bpexpdate each day is not something I am looking forward to.
Could be done by batchjobs and scheduled tasks but is too complicated I
think.

Get the impression Netbackup cannot overwrite jobs on a tape. It just
appends jobs on tape 1, takes tape 2, if full takes tape 3 etc. When all
the jobs on tape 1 are over the retention period tape 1 is used again.
Anyone confirms this?

best wishes,
Marcel

-----Original Message-----
From: Jewell, Jason T [mailto:jj151001 AT exchange.daytonoh.ncr DOT com]
Sent: Wednesday, May 23, 2001 1:36 PM
To: veritas-bu AT mailman.eng.auburn DOT edu
Subject: RE: [Veritas-bu] overwrite tape


When we have frozen and/or expired tapes we erase them (magnet) and
delete
them from the media server and master server databases, then re-insert
them
as new media (we use bpexpdate to delete from media servers , then
vmquery
-deassignbyid and vmdelete to delete from master servers, but we don't
have
3.4 so it may be different).  They will then be "overwritten."  This may
be
more work than you are looking for.

That is one solution you could use without messing with retention
levels,
but we also would be interested in hearing other procedures.

Jason Jewell
NCR 1611 Main ST
Dayton OH 45479
937-445-4173


-----Original Message-----
From: Marcel van den Berg (Devote) [mailto:marcelb AT interxion DOT com]
Sent: Wednesday, May 23, 2001 6:13 AM
To: veritas-bu AT mailman.eng.auburn DOT edu
Subject: [Veritas-bu] overwrite tape


Hi,

Might be a simple question but could not find an answer in the archive
nor on the site of Veritas nor in the manuals.

We are using Netbackup Business Server 3.4.1 on Windows 2000 using a
Compaq tapelibrary. We have just one tape available for every day in the
week. So for every day there is a volumepool with one tape in it.

How  can Netbackup be configuered that it overwrites the tape at every
backup? Now it appends the job to the tape, resulting in messages
'unable to allocate new media for backup.... (96) after a few
successfull backup's

Is the only way to control this the use of retention levels?

best wishes
Marcel
_______________________________________________
Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
_______________________________________________
Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
_______________________________________________
Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu





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