Display Magnified Output 

Note 2069.0            Questions on DMO and NSR behaviour         1 Reply
JOBURG::DOTTI "Ernesto Dotti - SI DEC South Africa"  36 lines  29-FEB-1996 07:35
--------------------------------------------------------------------------------
    I have a few questions concerning NSR-DMO coming from out customer who
    has been sort of participating to the external field test of DMO and
    that has started running the EFT version with the official oracle
    release of PBR.
                                          
    Here below is the text of the e-mail as sent by the customer.
    
    Thanks in advance. Best regards, Ernesto.
                                             
We have everything installed and we managed to perform a backup, but there are
a few things that I am unsure about and I would like you to investigate them.

In the DMO manual it explains that you can use the command-line (obackup) or
the Networker GUI (nwadmin) to perform backups. It also gives you step by step
instructions on how to set up NSR for DMO.

I have done all this according to the manual. I have a Directive called Oracle
which uses obkasm, A schedule called Oracle, A pool called Oracle and one Group
called Oracle which consists of just the one machine. This machine is also the
Networker Server.

I have used the command-line to backup a database (obackup <script file>). It
works fine, but how do I specify which pool of tapes it should use? It uses the
Default pool and I want it to use the Oracle pool.

When using the GUI (nwadmin), you have to specify the Oracle backup script as
the fileset it should backup. I have done it, but instead of backing up the
database, it just backs up the script file itself. Now it uses the correct tape
pool (Oracle).

According to the manual you can create a nsr.cf file with variable declarations
in it which will be read when the backup is started. This file must be in the
directory where obackup is called from. I have done so, but the file is not
read. I still have to declare a variable like OBK_HOME from the command-line.

When is the final version NSR_DMO going to be released?
================================================================================
Note 2069.1            Questions on DMO and NSR behaviour             
JOBURG::DOTTI "Ernesto Dotti - SI DEC South Africa" 102 lines  29-FEB-1996 07:56
                              -< Replied by mail >-
--------------------------------------------------------------------------------
        Thanks a lot for the reply, posted here for other's reference.
    
        Best regards, Ernesto
    
From:   DECWET::MORRIS "Terry Morris, SMS/NetWorker Engineering - DTN 548-8783  29-Feb-1996 0751" 29-FEB-1996 17:49:54.88
To:     JOBURG::DOTTI
CC:     MORRIS
Subj:   RE: Questions and support for NSR DMO

    Ernesto,

> I have used the command-line to backup a database (obackup <script file>).
> It works fine, but how do I specify which pool of tapes it should use?
> It uses the Default pool and I want it to use the Oracle pool.

Set the NSR_GROUP environment variable to the name of the savegroup "Oracle".
Then the save that is started by obackup will be considered to be in that
savegroup, and will use its pool.

> I have a Directive called Oracle which uses obkasm, A schedule called Oracle,
> A pool called Oracle and one Group
> called Oracle which consists of just the one machine. This machine is also the
> Networker Server.

> When using the GUI (nwadmin), you have to specify the Oracle backup script as
> the fileset it should backup. I have done it, but instead of backing up the
> database, it just backs up the script file itself.

There are several possibilities.  First, it may be that obkasm is not being
used.  You can tell by watching what processes are started on the database
server machine (which, as you've said, is also the NetWorker server) when the
savegroup is started.  To see the processes, start up a terminal window,
and after you start the savegroup, type "ps axwww".  Repeat this while the
savegroup runs, and watch for processes running "savefs", "save", "obkasm",
"obackup", and "nsrobkcon".  If you see only "save" or "savefs", then obkasm
isn't being run.  If you see obkasm, but not obackup or nsrobkcon, then
obkasm may have had trouble accessing its configuration files or the obackup
script file.  Let us know what processes you see.

If obkasm isn't being run, then it may be that the NSR client setup for
the database machine is missing the Oracle directive.  To check this, start
nwadmin, open a "Client setup" window, select the database machine, and look
at the "Directive" box.  It should have the Oracle directive.

(Note that later you may want to have a more general directive that can handle
both regular files and database files.  We can talk about this after you have
a successful scheduled save that backs up the database.)

Could you also please post the contents of your Oracle directive, if it's
different from what's in the DMO Admin Guide?

> According to the manual you can create a nsr.cf file with variable
> declarations
> in it which will be read when the backup is started. This file must be in the
> directory where obackup is called from. I have done so, but the file is not
> read. I still have to declare a variable like OBK_HOME from the command-line.

There are several different kinds of files used with obkasm and obackup,
which go in different places -- it's easy to miss this distinction:

The .cf files go in the same directory as the obackup scripts that they're
associated with.  You can have an nsr.cf file, which will be read first and
used for every obackup script in that directory, and, if you want, a .cf file
for each script with options that apply only to that script.  These
script-specific files should have names that are the name of the obackup
script but with .cf added to the end.

The files that must go in the same directory as obkasm are the precmd and
postcmd scripts.  Recall that "precmd" and "postcmd" are lines that go in
the .cf files, and point to scripts that obkasm will run before (precmd)
and after (postcmd) it runs obackup.  The actual scripts that precmd and
postcmd specify must have names ending in .obk, and must be in the same
directory as obkasm, which is by default /bin.  (This is for security reasons:
These scripts are run as root, so if some unscrupulous person were able to
insert a command, they could do whatever they pleased on the system.  Placing
these files in /bin, which has standard protections that don't allow non-root
users to modify files, makes is more likely that they will be safe.  The same
caution applies to *any* script that's run as root by some automatic procedure:
root's crontab entries, root's .login, any startup scripts, etc., etc.  One
should always be careful that such things are not writeable by non-root users.)

So, there are three different kinds of files:

  - obackup scripts, which contain commands to obackup (see the Oracle Backup
    Utility manual for more detailed examples of these) -- these go in whatever
    directory you want (both the DMO and Oracle Backup Utility manuals suggest
    having a special script directory)

  - DMO configuration (.cf) files -- these go in the same directory as the
    obackup script they're used with

  - shell scripts invoked by precmd and postcmd -- these go in the directory
    where obkasm is (probably /bin).

It is possible that without some environment variable set, obkasm cannot
successfully start obackup, so if obackup can't find its .cf files, this could
be the cause of the problem.

Let us know if any of these things are helpful.

\tm
    
================================================================================