copy index's from 1 server to another

[xtrac]

Folks,

Can i copy the index files from 1 legato server to another if they are using the exact same version, or do i need to put it to tape and then mmrecov? if so it there a better white paper on how to do this than the one on legato's site?

Regards

-Roy

 

[john7]

Are you trying to merge two servers together? If you are no. It will not work. The media databases can not be merged that easily. If you are talking about upgrading your hardware to another system then the safest and best way to do this is do a full backup of you indexes and use mmrecov.

to get a full backup of your indexes to one tape use a this command

savegrp -0 -l full groupname

if all your clients are in the one group you will get a full backup of all indexes, if they are not use this command for all groups and you will have a full of your indexes. This command only backups up the indexes it does not backup any data from the clients.

John7

 

[soder]

Hmmmm.... I cant see why this shouldn't work... but why do you want to do that... why would you like to browse one clients index if it's impossible to recover the data.The data is located on the other servers backup tapes....

If you would like to move the server to a new one... use mmrecov...

 

[xtrac]

Thanks Guys,

I'm not trying to merge, but i cant get mmrecov to work, it recovers the media db, but won't recover the server index, thats why i thought i might just copy them over. I'll keep pluging away, thanks for your help.

-Roy

 

[fpastano]

The mmrecov command simple recover the media db and the res directory.
To recover the index you must use the nsrck -L7 command.
But is more easy to simple put down the legato services, move the empty index directory to index.OLD and copy your index dir.
Restart the services and all will work fine (it is good also for the mm dir, but not for the res cause you must edit some field to let this one work)

Bye

 

[eskolnik]

I am trying to do almost the same thing
I am planning on moving my Networker server from one machine to another and was wondering if you can point me in the right direction
The current backup server :
Sun 220R
Sun 450 MHz
512 MB RAM
two 16 GB hard drives (not mirrored)
100 Mbs Ethernet adapter
Sun L280 Auto Loader (6 slots) with one DLT7000 tape drive
Solaris 8
Networker 6.1.2 Build 340

/nsr is in the root patrition on this machine

The new backup server will be:
Sun Sun Fire 280R
twin 900 MHz processors
2 GB RAM
two 36 GB hard drives (Mirrored)
100 Mbs Ethernet adapter (for now)
twin Gigabit Ethernet adapters for later
STK L40 tape library (40 slots) with three SDLT tape drives
Solaris 9
Networker 6.1.2 Build 340

/nsr (link) actually in /data01/nsr

The mmrecov wont restore the res files...... Anyone else....

Ed

 

[Chapter11]

eskolnik: moving Networker from one server to another is easy, much easier than most people give it credit for.

Assuming you are keeping the original pathing...:

On the original server, tar up the entire networker installation (for me, it's in /sprd/legato).

On the new server, perform a normal server installation of Networker of the same version you've got running, then just untar the archive you made over the installation. No mucking around with mmrecov, no mucking around with index recovery. Yes, it works, I did it.

The only real issue is the license: you might have to contact your vendor to transfer the license to the new machine and obtain new enabler codes, but Networker should continue functioning for a few days. At the worst, you may need to send your nsr.res file to your vendor for them to hack on.

 

[gdc2001]

mmrecov restores the media database to the directory location of that install of networker:

Old machines install path

/opt/nsr/mm

New machines install path

/nsr/mm

So a an mmrecov run on the new machine will place the media database in the New machines mm folder, however the res.r folder that is your recovered res is restored to its original location. In the case of this example, on the new machine the res.r will restore to /opt/ner/res.r

After you have moved the res.r to the correct location and renamed it you can run an nsrck -L7 to recover the indexes, again these go back to their old location and need to be moved by hand.

However if the insall path is the same on both machines then Chapter11's solution will work a treat.

 

[wallace88]

You can tar up the index directories and then move them with ftp or copy. Make sure that you copy over all the indexes, since the index path can be changed by the user.

The res directory can be also copied too. Note that the files in the res directory are not text files, even though they look like it. There is actually a checksum bit embedded into the file so that if corruption occurs, then it will be detected. This is why editing the res files *could* lead to corruption.

The media database should only be moved by backing up and recovering the bootstrap. In fact, in NetWorker 6.2, this applies even for the res directory. The media database is in a WISS database format that does not compress or copy well.

Now you can try simply tar'ing the mm directory up and relocating it, but how do you know for sure that the database is sound after the move? NetWorker may come up, but the corruption may not be detected at all. Why would you take the chance, when it's a simple matter of backing up and restoring the bootstrap?

 

[Chapter11]

wallace88: How does a file "not copy well"?

The media database in 6.0 is the same as 6.1 and 6.2, and it tars up and copies just fine.

If you want to change installation directories, naturally that's a different trick, but if you're going from one machine to another and using the same directories, there is no reason any file should "not copy well"

 

[wallace88]

The media database is in a special database format that has "spaces" (called place holders) in the database file structure. When these files are copied, usually these spaces are squeezed out; i.e. if the last part of the database is only spaces, the copy/zip/whatever may not back that part up.

So, like I said, copying over the media db *may* seem to work. Are you going to take a chance on important data? Months down the road when you need to recover data, and then NetWorker queries the media database to find what volume to load, only to read the part that became corrupted and then return the message "volume unknown"?

If chapter11 says it works for him, then good for him, but when it comes time for you to recover and then find out you can't... well you're going to have to go back to a bootstrap than may no longer exist, and then you'll have to start scanning tapes again.

I think backing up and restoring the bootstrap is a much simpler process. At least the much larger indexes can be copied over, but only if you're using NetWorker 6.x and above.

 

[coachcoty]

I had just done this. I built a new LEgato server and had a couple of client indexes that i needed to bring over. If this is Unix tar the indexes up and FTP them over and untar in your index directory. Define the client name to the server. Run nsrck -L6 clientname. This will recover the indexes. Now if you need to get the data from the tapes over you will have to either scan the whole tape or scan for the specific client. You will than need to add and inventory the tape into Networker. Hope this helps.

Ed