Migration of Master and media servers

[ram619]

Hi All,

 We are having two backup location EDC & GPT.

We are going to shut down GPT  location which consist of 1 master server (W2K3-32BIT) & 2media server’s (W2K3-32BIT) 1 tape library with 4 tape drive’s.

And the backup images resides in GPT location are with infinity retention.

We will move the above mentioned hardware’s to EDC location.

We need a pre-request activity step’s to shut down the GPT location. And after the migration if we get any restoration request from the old GPT location. How we will trace the Media Id information. Since we are going to use GPT location Master server as a media server in EDC and the possible way of doing the GPT backup image restoration in EDC location.

[Sebastian_Baszc]

Hey Mate,

You have 2 choices. Easy way and a hard way. Easy way is most costly, though.

Easy way: Merge NetBackup domains. To accomplish this you need to Symantec Representative and ask for the consulting services as Symantec exclusively has the tools to extract information from the one catalog and merge this information with the NetBackup Catalog/Databases in your Primary location.

Hard way: Take all tapes you have in GPT and import them in EDC. It can take some time, depending how many tapes you have and how many availabe tape drives you have in EDC. Once done, shutdown Master Server in GPT, move the hardware to the EDC and reinstall it as the Media Server with the original name.

Hope it helps

Seb

[RiaanBadenhorst · Accepted solution]

Hi,

There is a third way which is also qiute cheap. Its called recovery without import. What you basically do is copy all the catalog images from the GPT master to the EDC master. Then you take all the tapes from GPT and stick them in the library at EDC (or keep them handy for when you need them) but place them in a pool called FROM_GPT (or something like that) and inventory so the EDC master knows about them.

Because the images are in the catalog you'll be able to browse them using BAR, and if you restore the media will be found as the EDC master was told about them (with the inventory).

You'll have to make sure that there are no duplicates between the sites though i.e servers with the same name, media with the same id etc.

This process is usually used when you want to do a DR without having to do catalog restore, or a DR in an evironment that is running a separate Netbackup domain.

The media in your FROM_GPT pool would therefore not have an expiry as such as the EDC master doesn't have a record of it ever writing to them. They should be set to write-protected as well to ensure it doesn't try to. You might also need to note down the media from GPT's expiry time so you can "release" them into the EDC pools when they've really expired.

See the attached document.

I'm going to ask my colleague Marianne to also look at this response so I'm sure I've not missed anyhting :)

[Marianne]

I definitely vote for Recovery Without Import method.

Since this method is designed for DR, you just need to be extra careful when implementing it in production.

Additional steps that I will throw in:

While GPT master is still up and running, add the EDS master as SERVER to all clients.
Once test backups have completed successfully, you can use Host Properties, select EDS -> Make Master

For media servers - it will probably be easiest to remove NBU and reinstall instead of trying to update the Master and EMM server entries. Use the Device wizard on EDS master to scan and add devices for these media servers.  Keep GPT master around until all clients are working fine with new master. Remove NBU from GPT master and install as media server.

Extract a list of Assigned media on GPT master. Use vmadd on EDC master to add these tapes to the FROM_GPT pool. or maybe call the pool GPT_RESTORE.

All the unassigned tape can simply go to the Scratch pool.

**** EDIT *****

The more I think about this, the more steps come to mind...

Policies: Depending on the amount of policies on GPT master, you might also want copy over the class folder under netbackup\db when copying the images folder. You will HAVE to remember to update all STU's in policies afterwards.

Since the GPT images have Infinite expiration, you will probably want to fix the Unassigned state of the media. Run 'bpimport -create_db_info -id <mediaid>' for each of the tapes in GPT_RESTORE / FROM_GPT pool.

**** NOW I need Riaan, Bob and Andy to all look at the stuff that I've added !  This is NOT for the feint-hearted, probably best to pay a knowledge person at a Symantec reseller... and even better if you can test it in a lab first ****

[RiaanBadenhorst]

Hi,

Sorry I didn't add any info relating to the migration of the current cliets to EDC. I agree with Marianne, the steps listed in her post seem accurate.

One more thing to take care of are your database scripts. The master server name is listed in all SQL scripts so you'll need to update or recreate them. There could also be references to the master in scripts for Oracle and other database should you have any.

If you have any type of vault configuration i would note down the settings while the GPT master is still online. Recreate in the EDC domain. Do you have any other options configured in GPT, BMR, SLP, etc etc?