I think your only option to get a backup from master1 to master2 is to
import the tapes from master1 into master2. I'd be sending duplicates to
the master2 site and if disaster ever hits, you then import them into
master2.
Duplicating between master servers (bpduplicate kind of duplication) isn't
possible.
Any Master-of-master owners out there? Is it possible with a MoM?
-M
-----Original Message-----
From: Ellwood, MW (Mike) [mailto:M.W.Ellwood AT rl.ac DOT uk]
Sent: Friday, October 31, 2003 9:07 AM
To: veritas-bu AT mailman.eng.auburn DOT edu
Subject: [Veritas-bu] Want to bpduplicate across Master Servers
[Netbackup 3.4.1GA on Solaris platforms, to Sun L20 robots with LTO drives]
Hello,
We have a requirement to copy our main production backup overnight to
another almost
identical system running in another building. This is actually our
development system
which is a potential fallback system in the event of major outages in the
main building.
We would like a copy of the production backup away from the main building
(without
having to physically remove and take it each day!).
Our production host is pacioli1 and the development/fallback host is
pacioli2.
At the moment, each has its own Netbackup catalogue and totally separate set
of tapes.
i.e. pacioli1 is its own Master and Media server, and so is pacioli2.
I read in one of the manuals (sorry, no reference to hand) that in order to
do the kind
of thing I was intending, I would need to combine them to have a single
catalogue, so
presumably, for example, I would make pacioli1 the Master Server, with both
pacioli1 and pacioli2
being Media Servers, or vice-versa.
When I discusses this with my boss, he quite reasonably pointed out that in
a way, this
would defeat the object, since if we lost pacioli1 in some disaster, we
couldn't run
Netbackup on pacioli2 without setting it up as the Master Server, which
would presumably
require a lot of work, including presumably restoring the caalogue to
pacioli2. At least
with the present setup, if pacioli1 disappears we can run Netbackup on
pacioli2 "as is",
and can restore the pacioli1 data to pacioli2 providing we have suitable
tapes available.
To that end, and as a stop-gap until we can to the kind of duplication we
would really like,
I do a second backup on pacioli2 of pacioli1 as a client, so we always have
backups of
pacioli1 in the pacioli2 robot (or the firesafe in its building). For good
measure, I also
do a bpduplicate on pacioli1 to one of its own tapes, and store those
separately.
However, I felt sure that somehow, we should be able to do what we want, and
from
time to time, I've experimented with various options. The interesting thing
is that
I can definitely talk from one system to the other's tape robot, and we make
use
of this by restoring into pacioli2 direct from pacioli1's robot, in order to
create
development versions of the production system. I had to make judicious
additions to
bp.conf and vm.conf in order to get this to work, but work it definitely
does.
And of course, it's a cinch to backup pacioli1 from pacioli2 - it's just
another client.
I've just been experimenting with bpduplicate again, "determined to crack it
this time", but
it's still defeating me. Major features:
in my bpduplicate, running on pacioli1, I specify
-dstunit pacioli2-hcart-robot-tld-0
-M pacioli2
-c pacioli1
-shost pacioli1
In the first part of the log, it looks like it's going to work, since it
seem to be able to identify
the backup image and the tape - sample snip from log:
Duplicate started Fri Oct 31 11:02:42 2003
11:02:44 INF - Destination storage unit pacioli2-hcart-robot-tld-0 on host
pacioli2.admin.rl.ac
.uk.
11:02:45 INF - Duplicating class mwetest, schedule mwetest_s
(pacioli1_1067519734) created 10/3
0/2003 13:15:34, media id 8444L1.
11:02:46 INF - write host pacioli2.admin.rl.ac.uk: no entity was found (227)
11:02:48 INF - Duplicate of backup id pacioli1_1067519734 failed, no entity
was found (227).
11:02:49 INF - Status = no images were successfully processed.
If I do a bperror on pacioli2, there's a bit more info:
10/31/2003 11:15:55 pacioli2.admin.rl.ac.uk pacioli1 db_getIMAGE() failed:
no
entity was found (227)
...and this seems fairly consistent. I've tried running it on pacioli2 as
well, with identical results.
(The error shown in bperror is seen on pacioli2, whether I run it on
pacioli1 or 2, which is
sort of logical, I guess - I am asking pacioli2 to read the image and wwrite
it to its tapes
in both cases).
Preview (-p or -PD or -PM) run on either machine shows the correct image -
it's only when
I come to run it for real that I get the error - which makes me wonder if
it's actually a bug.
For the first time today, I thought of turning on the activity monitor, for
admin and dbdbm.
This shows a bptm command being built which looks like it has all the right
things
(is bptm documented? - perhaps it's only an internal command):
[pacioli1 admin log]
14:01:45 [27458] <4> build_write_cmd: /usr/openv/netbackup/bin/bptm bptm
-dup -cmd -nosig -v -L /tmp/run.tapecopy.mwetest.log -ru root -rclnt
pacioli1 -mediasvr pacioli2.admin.rl.ac.uk -c pacioli1 -b
pacioli1_1067519734 -cl mwetest -sl mwetest_s -bt 1067519734 -st 0 -rl 3 -ct
0 -cj 2 -den 6 -rt 8 -rn 0 -p week2_prod -date 1070198134 -cn 2
it continues:
14:01:45 [27458] <2> getsockconnected: host=pacioli2.admin.rl.ac.uk
service=bpcd address=130.24
6.102.55 protocol=tcp reserved port=13782
14:01:45 [27458] <2> getsockconnected: Connect to pacioli2.admin.rl.ac.uk on
port 935
14:01:45 [27458] <2> start_command: received BPCD success message for fork
request
14:01:46 [27458] <2> dupreadline: read: EXIT STATUS 227
There doesn't appear to be any corresponding activity log entry on pacioli2.
So, is it really impossible, or is there a way to do this?
Thanks!
Mike Ellwood
_______________________________________________
Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
|
