Conference decwet::networker

is the script placed in /usr/bin (wherever save is)?
is it chmod/chown properly?

what happens that is wrong?

    Hi
    
    The script is worink ok, the customer only wants to make sure
    that using savefs directly from a script does not cause him 
    any problems when he wants to restore something.
    (The man page for savefs sas that it is not recomende to
    use savefs directley ....)
    
    
    KMN

using savefs "directly" will do exactly what it is advertised to do: save
 a filesystem to the NetWorker server.

As files are saved, the info about the file will be logged in the index for the
 client doing the savefs. Recovers will read the index first, and allow the
 user to select desired files, and perform the recovery.

The "gotcha" is that savefs will not cause the server to save the *index file*
 for that client, thus, were *anything* to happen to the index file, recovers
 would not be possible, since you cannot copy those index files back from a
 save tape!

FYI: this is why savegroup is used.

    Hi
    
    Would it help to use save insted of savefs ???
    
    
    
    KMN

actually, the job of savefs is more of "administration" between savegroup
 and save. It responds to the server (savegroup) about those filesystems
 that are "local" and need saving. At a later point in time (a few seconds
 usually), the server then tells savefs to start a save on a filesystem,
 which causes it to exec a save process.

So, no, using save directly won't change the issue about the indexes being
 saved.

I must be missing something here.

If the concern is that save/savefs does not "automatically" save the
server's indexes, then why not just (explicitly) save the server's indexes?

Set up a group which contains only the server which is "disabled"
(i.e. the whole savegroup will not be automatically started by NetWorker)

Then, on whatever schedule you wish, run
savegrp -O <group-name>

savegrp -O causes savegrp to save only the indexes for any client in the
group.  If it is the server, then the bootstrap save is done.

This is what you want, right?

Kevin

For the script in .0, what saveset(s) are in the client resource?

Wes

Never mind.  I was having a problem making the script work, but I 
understand now.

Wes

    Hi
    
    The script in .0 is a pre/post script, so it is "started" from the
    networker server which in this case is atno50, and the client is
    stkh02. When running backup (using the script in .0) on stkh02 the
    messages on the servere is this:
    
    
    stkh02.alcatel.no: /              level=incr,     10 KB 00:00:15
    8 files
      stkh02.alcatel.no: /usr           level=incr,    241 KB 00:00:38
    7 files
      stkh02.alcatel.no: /var           level=incr,    123 KB 00:00:07
    7 files
      stkh02.alcatel.no: /home          level=incr,      4 KB 00:00:06
    3 files
      stkh02.alcatel.no: /opt           level=incr,      0 KB 00:00:07
    0 files
      stkh02.alcatel.no: /var/opt/openmail level=incr, 1.9 MB 00:00:12
    31 files
      stkh02.alcatel.no: /tmp           level=incr,      3 KB 00:00:05
    2 files
      stkh02.alcatel.no: /work          level=incr,      0 KB 00:00:04
    0 files
      atno50.alcatel.no: /index/nsr_031096/index/stkh02.alcatel.no level=9,
    170 KB 00:00:29      2 files
    
    
    It seems to mee that indexes is beeing saved, does't it ???
    
    
    
    Regards
    
    Kjell-Magne Nilstad
    DIGITAL Norway

    Hi
    
    Sorry to belabor this, but I need to get clearer on two points:
    
    o When I use a pre-/post-processing script, should I be using save or
    savefs in that script?
    o will either the save or savefs command save the index files, or do I
    need to do this manually somehow as part of my script?
    
    Thanks!
    tl

    o When I use a pre-/post-processing script, should I be using save or
    savefs in that script?

 save

    o will either the save or savefs command save the index files, or do I
    need to do this manually somehow as part of my script?

 you need to save the index manually. At least, according to what I know 
 (kinda from Legato, etc), although Kjell-Magne Nilstad seems to have
 demonstrated otherwise.

   Hmmm, guess it's time to play with NSR V4.3...

root@decaf[2]> emacs save-bwe-pre
root@decaf[3]> chmod 777 save-bwe-pre
root@decaf[4]> cat save-bwe-pre
#!/bin/sh
PATH=/sbin:/usr/sbin:/usr/bin
date >> /tmp/bwe-pre.txt
echo "Starting save pre-processing" >> /tmp/bwe-pre.txt
echo $* >> /tmp/bwe-pre.txt
save $*
echo "Completed save pre-processing" >> /tmp/bwe-pre.txt
date >> /tmp/bwe-pre.txt

-----------------------
 In server GUI, changed client resource for decaf to have "save-bwe-pre"
 in the backup command field, then APPLY.
-----------------------
 client is decaf, in group Default, not enabled
 server is mrcofe, in group "prod", not enabled
 there are 2 mount points: /, and /usr, on decaf
-----------------------
 in GUI, Group Control, start now on group Default....

root@decaf[18]> cat bwe-pre.txt
Fri Apr 11 15:06:33 PDT 1997
Starting save pre-processing
-s mrcofe.zso.dec.com -g Default -LL -f - -m decaf.zso.dec.com -t 860796312 -l i
ncr -q -W78 -N /usr /usr
Completed save pre-processing
Fri Apr 11 15:07:13 PDT 1997
Fri Apr 11 15:07:19 PDT 1997
Starting save pre-processing
-s mrcofe.zso.dec.com -g Default -LL -f - -m decaf.zso.dec.com -t 860795196 -l i
ncr -q -W78 -N / /
Completed save pre-processing
Fri Apr 11 15:07:41 PDT 1997

  (in GUI, /usr was saved first, then /, and finally index for decaf, then
  index for mrcofe)
-----------------------------
 in GUI, alter group "prod" to be enabled, thus placing server into an
 enabled group.

-----------------------------

Fri Apr 11 15:10:00 PDT 1997
Starting save pre-processing
-s mrcofe.zso.dec.com -g Default -LL -f - -m decaf.zso.dec.com -t 860796393 -l i
ncr -q -W78 -N /usr /usr
Completed save pre-processing
Fri Apr 11 15:10:39 PDT 1997
Fri Apr 11 15:10:40 PDT 1997
Starting save pre-processing
-s mrcofe.zso.dec.com -g Default -LL -f - -m decaf.zso.dec.com -t 860796439 -l i
ncr -q -W78 -N / /
Completed save pre-processing
Fri Apr 11 15:11:07 PDT 1997

  (in GUI, /usr was saved first, then /, and finally index for decaf)

    > will either the save or savefs command save the index files, or do I
    > need to do this manually somehow as part of my script?
    
    >> you need to save the index manually. At least, according to what I know
    >> (kinda from Legato, etc), although Kjell-Magne Nilstad seems to have
    >> demonstrated otherwise.
    
    Bruce, I think the index is saved at the initiative of the savegroup.
    So, as long as the "save_bwe_pre" command is initiated from the savegrp
    command, the index of the client should be saved. And concerning the
    index of the NSR server, it depends of the fact that the SERVER client
    is in a "enabled" group.
    
    Am I in trouble ?
    
    Manu.

You've got it right.

If the server is part of a savegroup which is enabled (scheduled) for normal
saves, the server's indexes (bootstrap) will be saved as part of that savegroup.

If the server is NOT part of any enabled savegroup, the bootstrap save
will be done any time ANY savegrp is run.

savegrp -o <group-with-server-in-it> is the way to manually trigger saving of
the server's indexes (bootstrap)

If savegrp is never run, the server's indexes are never saved.

Clear?

Kevin

    Hi
    
    I'm using a pre-processing script, and attempting to save to a group
    with the name "Testing new clients".  It seems that the save command
    can not handle group names with spaces in it.
    
    Here's how I'm testing it; I saw that this was the actual command
    passed by NetWorker Server V4.2B to this Digital UNIX client:
    ksh -v 'nsr_save_test -s nw12780.hlo.dec.com -g "Testing new clients"
    -LL -f - -m email1.hlo.dec.com -l full -W78 -N /usr/users /usr/user
    
    Here's the output of the save command:
    save $*
    + save -s nw12780.hlo.dec.com -g Testing new clients -LL -f - -m
    email1.hlo.dec.com -l full -W78 -N /usr/users /usr/users
    new: No such file or directory
    clients: No such file or directory
    -LL: No such file or directory
    -f: No such file or directory
    -: No such file or directory
    -m: No such file or directory
    email1.hlo.dec.com: No such file or directory
    -l: No such file or directory
    full: No such file or directory
    -W78: No such file or directory
    -N: No such file or directory
    save: SYSTEM error, no group `Testing' configured on server
    save: Cannot open save session with nw12780.hlo.dec.com
    #
    
    Apparently, since the -g value of 'Testing new clients' is not quoted,
    save is getting confused, and the command fails.  Any thoughts on this?
    
    Thanks!
    tl

    Hi
    
    Are there any other things I can test?  Am I doing something obviously
    wrong?
    
    Thanks!
    tl

    It is late for me (10:35pm) and I try to remember, but is $@ not
    keeping the argument list ?
    
    Thus, maybe you can try " save $@ " ?
    
    Manu.

This week has been pretty full so far.  I tried duplicating the
 situation and was able to. it appears that savegroup is sending
 down a quoted string for the group name, and it comes into a shell
 as 1 "word" (see man csh, or sh for definition of "word"), but when
 I got into trying to make the script execute the save command
 using that "word", I discovered it simply comes down to getting that
 string parsed by the shell as 1 word, not as multiple words.

in english: if <foo abr bletch> cames in, then you *need* to send
 out a command that looks like

	save ... -g 'foo abr bletch' ... /path

 or not-nice-things happen. So far, I cannot think of a way using csh, sh,
 or ksh to do this, since they usually use single quotes (') as a magic
 character.  Perhaps Perl....

Bruce

    Hi Bruce
    
    First off, I really appreciate the time you spent on this, as I know
    you folks are very busy.
    
    But, but, why does save work everynight in my environment with group
    names that include spaces, and now won't work in a ksh shell script? 
    What's different about how save is called in those two cases?
    
    Thanks!
    tl

    Hi
    
    I spent another hour on this myself.  It gets weirder and weirder.  I
    modified my script to, instead of using 'save $@', I explicitly
    specified all of the needed arguments on the command line; that way,
    Ithought I'd get the command shell out of the business of parsing the
    command line values.  Also, I put quotes around the group name value
    before I fed it to save.
    
    It still failed:
    
    email1.hlo# ksh -vx nsr_save_test -s nw12780.hlo.dec.com -g "Testing
    new clients" -LL -f - -m email1.hlo.dec.com -l full -W78 -N /usr/users
    /usr/users
    #!/bin/ksh
    #
    # -s mrcofe.zso.dec.com -g Default -LL -f - -m decaf.zso.dec.com
    #  t 860796312 -l incr -q -W78 -N /usr /usr
    # EDIT THE FOLLOWING VARIABLE WITH THE VALUE OF THE PARTITION
    CONTAINING
    # THE MAILWORKS USER AREA.
    #
    MAILWORKS_USER_AREA=/usr/users
    + MAILWORKS_USER_AREA=/usr/users
    #
    # Take the arguments passed to the command procedure, and places the
    ones
    # that are useful into variables.
    #
    echo "Just before the getopts"
    + echo Just before the getopts
    Just before the getopts
    while getopts s:g:L:f:m:t:l:qW:N: opt
    do
       case $opt in
          s)   SERVER=$OPTARG;;
          g)   GROUP='"'$OPTARG'"';;
          f)   FILE=$OPTARG;;
          t)   TIME=$OPTARG;;
          l)   LEVEL=$OPTARG;;
          W)   WIDTH=$OPTARG;;
          N)   NAME=$OPTARG;;
          m)   ;;
          L)   ;;
          q)   ;;
          *)   echo "Error: $opt $OPTARG";;
       esac
    done
    + getopts s:g:L:f:m:t:l:qW:N: opt
    + SERVER=nw12780.hlo.dec.com
    + getopts s:g:L:f:m:t:l:qW:N: opt
    + GROUP="Testing new clients"
    + getopts s:g:L:f:m:t:l:qW:N: opt
    + getopts s:g:L:f:m:t:l:qW:N: opt
    + FILE=-
    + getopts s:g:L:f:m:t:l:qW:N: opt
    + getopts s:g:L:f:m:t:l:qW:N: opt
    + LEVEL=full
    + getopts s:g:L:f:m:t:l:qW:N: opt
    + WIDTH=78
    + getopts s:g:L:f:m:t:l:qW:N: opt
    + NAME=/usr/users
    + getopts s:g:L:f:m:t:l:qW:N: opt
    #
    # If this execution is saving the MAILworks user area, then shutdown
    MAILworks
    #
    if [[ $LEVEL = full && $MAILWORKS_USER_AREA = $NAME ]]
      then
      #/usr/opt/DMW/bin/stopMAILworks
      echo 'Stopping MAILworks Server'
    fi
    + [[ full = full ]]
    + [[ /usr/users = /usr/users ]]
    + echo Stopping MAILworks Server
    Stopping MAILworks Server
    #
    # Execute the NetWorker save command to backup the saveset [=partition]
    to the
    # NetWorker server.
    #
    #save $*
    save -s $SERVER -g $GROUP -LL -f $FILE -t $TIME -l $LEVEL -W $WIDTH -N
    $NAME
    + save -s nw12780.hlo.dec.com -g "Testing new clients" -LL -f - -t -l
    full -W 78 -N /usr/users
    new: No such file or directory
    clients": No such file or directory
    -LL: No such file or directory
    -f: No such file or directory
    -: No such file or directory
    save: invalid time specification: "-l"
    #
    #  If this execution is saving the MAILworks user area, then startup
    MAILworks
    if [[ $LEVEL = full && $MAILWORKS_USER_AREA = $NAME ]]
      then
      #/usr/opt/DMW/bin/startMAILworks
      echo 'Starting MAILworks Server'
    fi
    + [[ full = full ]]
    + [[ /usr/users = /usr/users ]]
    + echo Starting MAILworks Server
    Starting MAILworks Server
    # End of procedure nsr_save_mailworks
    email1.hlo#
    
    Note that even though I pass 'save -s nw12780.hlo.dec.com -g "Testing
    new clients" . . .', I still see errors from save that it is unable to
    correctly parse the group.  So, since this data is echoing correctly in
    ksh when the save command is processed, why can't save pick up the
    values correctly?
    
    Whats' different about calling save from a shell script that isn't done
    when save is called by NetWorker in the 'usual way'?
    
    Thanks very much!
    tl

    Hi,
    
    I think you misplaced the quotes ("). You placed it such that they are
    included in the variable $GROUP in the getopts routine:
    
    	 g)   GROUP='"'$OPTARG'"';;
    
    For me, you should remove it from that place and simply change the line 
    where you have the save from:
    
    	save -s $SERVER -g $GROUP -LL -f $FILE -t $TIME -l $LEVEL ......
    
    to:
    
    	save -s $SERVER -g "$GROUP" -LL -f $FILE -t $TIME -l $LEVEL ...... 
    
    
    This is my suggestion, 
    
    Best Regards, Manu.

    Hi Manu
    
    Yippee!  That was it!  When I quoted $GROUP on the save command line,
    save accepted the data:
    
    save -s $SERVER -g "$GROUP" -LL -f $FILE -l $LEVEL -W $WIDTH -N $NAME
    + save -s nw12780.hlo.dec.com -g Testing new clients -LL -f - -l full
    -W 78 -N /usr/users
    email1.hlo#
    
    So, it appears that using:
    
    save $@
    
    does not work, but using:
    
    save -s "$SERVER" -g "$GROUP" -LL -f "$FILE" -l "$LEVEL" -W "$WIDTH" -N
    "$NAME"
    
    will.  Is this a bug?
    
    Final questions:
    
    o is it a problem that I specify in my scripts the actual command
    attributes, rather than allowing 'save $@' to just figure it out?
    
    o what is the complete list of attributes that I should specify to save
    in these scripts?
    
    Thanks again!
    tl
    

    Happy to help you!
    
    Is it a bug ? - No, certainely not. It is true since the first bourne
    shell and accross all the other.
    
    For you other question, No you should not use $@, it was only a
    suggestion, and concerning the save, you do like you want.
    
    Manu.
    

    Hi
    
    The script presented in 485.0 contains the following:
    
    
    if [ $LEVEL = "full" ]; then
            savefs -s $SERVER -g $GROUP -l $LEVEL /
            savefs -s $SERVER -g $GROUP -l $LEVEL /usr
    .
    .
    else
            savefs -s $SERVER -g $GROUP -l $LEVEL -t $TIME /
            savefs -s $SERVER -g $GROUP -l $LEVEL -t $TIME /usr
    .
    .
    fi
    
    Is NOT specifying '-t {time}' a requirement for running a full backup,
    or will it work okay (ignore it) if specified on a 'save' command?
    I'd like to do without it if possible; less coding machinery, etc.
    
    Thanks!
    tl

    Thanks.  I interpret the two previous replies to mean that I can use
    the -t option on save when -l is 'full'.  True?
    
    Next question: in .0, I see:
    
    	while getopts s:g:L:f:m:t:l:qW:N: opt
    
    Is having 'qW', instead of 'q:W', a typo or a feature?  That is, should
    the command be:
    
        while getopts s:g:L:f:m:t:l:q:W:N: opt
    
    The above command is what my script has currently, and I'm noticing
    that save seems to report every file it saves (which is not what I
    want).
    
    Thanks!
    tl
    Thanks!
    tl

    Re .26 and .27
    
    When this command:
    
    -s nw12780.hlo.dec.com -g Testing new clients -LL -f - -m
    email1.hlo.dec.com -l full -q -W78 -N / /
    
    is fed to my command procedure, which contains:
    
    while getopts s:g:L:f:m:t:l:q:W:N: opt
    do
       case $opt in
          s)   SERVER=$OPTARG;;
          g)   GROUP=$OPTARG;;
          L)   ;;
          f)   FILE=$OPTARG;;
          m)   MASQUERADE=$OPTARG;;
          t)   TIME=$OPTARG;;
          l)   LEVEL=$OPTARG;;
          q)   ;;
          W)   WIDTH=$OPTARG;;
          N)   NAME=$OPTARG;;
          *)   echo "Error: $opt $OPTARG";;
       esac
    done
    
    save -s "$SERVER" -g "$GROUP" -LL -f "$FILE" -m "$MASQUERADE" -t
    "$TIME" -l "$LEVEL" -W "$WIDTH" -N "$NAME" "$NAME"
    
    I see this message in /logs/messages:
    
    Apr 24 20:13:18 robot1 crsupp: --- Successful Save Sets ---
    Apr 24 20:13:18 robot1 crsupp:
    Apr 24 20:13:18 robot1 crsupp: * email1.hlo.dec.com:/ save: saving
    files modifie
    d since Thu Apr 24 00:00:00 1997
    Apr 24 20:13:18 robot1 crsupp: * email1.hlo.dec.com:/ /dev/null
    Apr 24 20:13:18 robot1 crsupp: * email1.hlo.dec.com:/ /dev/tty
    Apr 24 20:13:18 robot1 crsupp: * email1.hlo.dec.com:/ /dev/ptyp1
    Apr 24 20:13:18 robot1 crsupp: * email1.hlo.dec.com:/ /dev/ttyp1
    Apr 24 20:13:18 robot1 crsupp: * email1.hlo.dec.com:/ /dev/ptyp3
    Apr 24 20:13:18 robot1 crsupp: * email1.hlo.dec.com:/ /dev/ttyp3
    Apr 24 20:13:18 robot1 crsupp: * email1.hlo.dec.com:/ /dev/
    Apr 24 20:13:18 robot1 crsupp: * email1.hlo.dec.com:/ /etc/ntp.drift
    Apr 24 20:13:18 robot1 crsupp: * email1.hlo.dec.com:/ /etc/
    Apr 24 20:13:18 robot1 crsupp: * email1.hlo.dec.com:/ /
    Apr 24 20:13:18 robot1 crsupp: * email1.hlo.dec.com:/
    Apr 24 20:13:18 robot1 crsupp:   email1.hlo.dec.com: / level=full, 11
    KB 00:00:1
    2     10 files
    
    I checked the file index, and only these 10 files were stored.  So it
    appears that:
    
    1) a full backup does not send a -t option to the save command
    2) the full backup assumes it should do an inc since the previous
    midnight, when fed '-t 0'
    
    So, it seems that two different save commands are required, one with -t
    (for incs) and one without the -t (for fulls).
    
    Thoughts?
    
    Thanks!
    tl

If a letter is followed by a :, that option is expected to have an argument.