Networker

Re: [Networker] Moving Networker server Solaris -> Linux?

2006-02-15 15:24:06
Subject: Re: [Networker] Moving Networker server Solaris -> Linux?
From: Matt Hemingway <matt.hemingway AT PCNALERT DOT COM>
To: NETWORKER AT LISTSERV.TEMPLE DOT EDU
Date: Wed, 15 Feb 2006 12:19:45 -0800
I've done this.  A search through the archives should pull up my
notes.  But here it is again:


============
Hi.

I just finished a migration from Networker 7.1.2 on Solaris 8 to Networker 
7.1.2 on Linux (Debian 3.0, Kernel 2.4.X).

The disaster recovery guide, provided by Legato, is what you want to follow 
(almost to a "T").

The first thing I did was stop Networker on Solaris, unplugged the network 
cable and hooked the tape library up to the Linux box.  I then installed Debian 
3.0.  Linux requirements, are 1. a supported SCSI card and 2. a kernel from the 
2.4.X series.  2.6.X does not work, I tried and had a ton of problems.  
Hopefully in the future it will be supported.  You'll also have less problems 
if you give/make the Linux box the same hostname/IP as the Solaris box was 
(hence unplugging the network cable).

I then took all the Networker RPM's and converted them to .deb's (alien -d 
$lgtopackage.rpm).  Then installed them all via dpkg -i.  The client and the 
server packages conflict with a file (networker.cluster, I think), so I had to 
add the dpkg -i --force-overwrite on the server package and it seemed to work 
fine.

Then, following the disaster recovery guide, started up Networker (nsrexecd && 
nsrd).  Then ran jbconfig and installed the drivers for our tape library/drive 
(a StorageTek L20 with a Quantum DLT8000 drive).  Then reset the changer (nsrjb 
-vHE).  Then did an inventory (nsrjb -I).

The next part is recovering the servers Bootstrap info.  I get e-mail after 
each backup with the latest bootstrap information.  I also took note of which 
slot the tape with the latest bootstrap is located in.  If you don't have this 
info available, then the disaster recovery guide will show you how to get it.  
Anywho, then I ran: nsrjb -lnv -S 3 -f /dev/nst0  which loads the tape with the 
latest bootstrap info into the drive.

Since I had all the bootstrap info and had the needed tape already mounted, I 
was able to skip the scanner -B command and moved right on to running mmrecov 
and entered in the correct info for the device, ID, file number, record number. 
 It did its thing and put the bootstrap info in /nsr/res.R.  I shutdown 
Networker (nsr_shutdown -a) and moved /nsr/res to /nsr/res.orig and then moved 
/nsr/res.R to /nsr/res

Now, since scsi device info is different on the Linux box, I change directories 
to /nsr/res and ran:

server:/nsr/res# rgrep scsidev *
nsrdb/05/005f13a13b328de2cf0da80c:control port: [email protected];
nsrdb/05/005f13a13b328de2cf0da80c:description: "[email protected]: Standard SCSI 
Jukebox";

On Linux it should be [email protected] (for me at least, when you run jbconfig 
above, it will tell you where your drive/library is located...all depending on 
the amount of SCSI devices you have and where they are located, I guess).  So I 
changed it to [email protected].  I then also created a symlink from /dev/rmt/0ubn 
to /dev/nst0.  /dev/rmt/0ubn was the device Networker was using when it was 
running on Solaris.  I didn't feel like grep'ing and changing the files, so 
just created the symlink.  Also, on the Solaris box, Networker was looking for 
logger and mail in /usr/ucb.  On Debian, those are both located in /usr/bin, so 
I just created another symlink from /usr/ucb/logger and mail to /usr/bin/logger 
and mail.  Remember, the /nsr/res directory is how it was on the Solaris box, 
so little things like this need to be taken care of when migrating to a 
different *nix platform.

Then, I started up Networker (nsrexecd && nsrd).  I had to wait about 5 minutes 
or so while the tape that was loaded last (the one I recovered the bootstrap 
info from) was read and mounted and shtuff.  Then I reset the autochanger 
(nsrjb -HE) and did an inventory (nsrjb -Iv).

Next step is to recover all the client indexes (nsrck -L7 client_name).  I took 
the latest e-mail I got from the latest group saves (solaris, linux and 
windows) and went down the list recovering all the client index files.

Then did a couple of tests (backup/recovery) and e-mailed Legato regarding 
getting new license keys/enabler codes.

I've only noticed one small problem so far.  The client indexes for our Windows 
hosts are located on our Offsite tapes.  I'm not exactly sure why they are 
written there, or why nsrck is wanting to grab them off of them.  We run 
Offsites on Friday nights and Onsite Fulls on Saturday.  Solaris and Linux 
client index recovers both read from the Onsite tapes.  Oh well, as long as 
they are recoverable...I guess.

Anywho, I think this covers it.  I've seen many posts to this mailing list 
(including ones from myself) regarding such a migration and any problems (that 
whole big and little endian thing seemed to be the main worry) but I didn't 
notice any...yet.  :)

Hope this helps someone.  Let me know if there's any questions or problems, I 
drank a wee bit last night and am not feeling 100% right now.  hehe.

-Matt

p.s.  At first I attempted the migration from Networker 6.1.something on 
Solaris to Networker 7.1.2 on Linux.  From what I can tell, it did work fine.  
The disaster recovery guide warns against this but I didn't seem to have any 
problems.  I then chickened out of that migration and upgraded the Solaris box 
to 7.1.2.


On Wed, 15 Feb 2006 14:00:26 -0600
Rich Graves <rgraves AT CARLETON DOT EDU> wrote:

> We're interested in moving our primary Networker server from an aging, 
> underpowered Sun box to Linux instead of replacing the Sun. The cost 
> argument is pretty strong -- a new Networker Network Edition for Linux 
> license costs about the same as annual maintenance on the existing 
> Solaris license, and the commodity hardware is cheaper. Other than the 
> server itself, the backup hardware is a few terabytes of FC disks for 
> staging (tier 2 D2D license) and FC-attached SDLT-320 drives for 
> cloning/archiving, all of which should work with a Linux server. I'm 
> aware that we could buy a Linux Storage Node, but that just delays the 
> problem.
> 
> I know that moving indices and media database between platforms is not 
> supported. Is it in fact impossible? Ever if just one can be moved, that 
> might be acceptable -- i.e., rescue the indices but need scanner to 
> actually read the tapes, or rescue the media database but lose indices 
> and need full saveset recovers to restore data.
> 
> Of course we could keep the Sun lying around and use a cheap standalone 
> drive for restores.
> 
> Who's been down this road before?
> 
> To sign off this list, send email to listserv AT listserv.temple DOT edu and 
> type "signoff networker" in the
> body of the email. Please write to networker-request AT listserv.temple DOT 
> edu if you have any problems
> wit this list. You can access the archives at 
> http://listserv.temple.edu/archives/networker.html or
> via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER

To sign off this list, send email to listserv AT listserv.temple DOT edu and 
type "signoff networker" in the
body of the email. Please write to networker-request AT listserv.temple DOT edu 
if you have any problems
wit this list. You can access the archives at 
http://listserv.temple.edu/archives/networker.html or
via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER

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