NBU 5240 - AIR and DR approach

[quebek]

Hello 

Scenario is as follows:

In site A is NBU 5240 appliance with MSDP configured making its own NBU doamin. All backups are AIRed to remote central site (DR site) by leveraging SLPs.

Site A goes down, data recovery is needed and NBU appliance there can not be used anylonger as it got also broken.

So we do ship new NBU 5240 to our DR  site and configure it as it was in terms of hostname, pools, etc and then what next?

How to have quickly most fresh backups AIRed to central location from site A on this new NBU 5240? Once these backups will be on that new appliance we will ship this appliance to site A to perform local restores... This new appliance will make its own NBU domain as it was earlier.

[D_Flood]

Is the DR site it's own NetBackup domain or is it just a Media Server of the primary domain?

The best case is that the DR site be it's own domain and, using the SLP Replicate function, you transfer the backups.  The SLP at the receiving end then imports the backup into the local catalog and it's ready to use within 30 minutes of the transfer being finished.

[quebek]

Hey

Yes the DR site is having its own NBU domain... And the backups were imported there via SLP from site A. I am not really following you? How may I apply a SLP to a already existing backup image?? As the backup images from site A are already available at DR site master server - so how to force or create a SLP to tell that it should replicate this and that backup image to the new NBU 5240 appliance?? Am I missing some magic on NBU end?

[D_Flood]

If the SLP on the prime domain is using the function "replicate" and the SLP on the DR domain is using the function "import" then nothing else is needed.

Once the import job runs on the DR domain master the backup is then in the DR domain catalog and you can restore it, work with it, etc just like any other backup.

Now it is true that getting the images back to the original domain after you've recovered the site is a lot more tricky but there is a KB about manual replication that, if you are wearing the right ruby slippers and it's the right phase of the moon, will work.

<soap box mode on>

But you may have just stumbled on one of the shortcomings of replication.  Unless you decided before the backup was run to apply the SLP to that backup and have the image(s) replicated, it's a <deleted> to manually replicate something after the fact.  It can be done but, like so many other features of NetBackup, the initial coding for replication was done to sell a version number and then never improved/fixed.  There's no reason, other than it's not a bell or whistle, to not be able to pick an image in the Catalog screen, right click on it, and select Replicate...oh well.

<soap box mode off>

[Mouse]

It's not possible to add images to SLPs retrospectively, however in AIR it is possible to replicate an image manually which is a bit tricky but sometimes it works http://www.veritas.com/docs/000017825

[quebek]

Hello Mouse

I saw this TN already and it is rather cumbersome, also I am not sure if it fits my scenario as "NBU master in site A was rebuilt from scratch"....

So it appears that there is no such option. This is sad...

[Mouse]

The technote is just an example how it could work. I think in a real 'DR' scenario you also likely to be interested in restoring NBU configuration which is not protected by A.I.R. - policies, SLPs, etc. These items are covered by the catalog backup and recovery process. 

Once you re-establish the new master, you then need to replicate images back one way or another. A.I.R.'s design geared towards replicating the backup image to one or multiple domains but there is no easy way to get it back to the source domain, this was one of the disadvantages of A.I.R. from the beginning.

As an alternative, you may want to look into design option with one domain and replicated images plus catalog backup/recovery if you don't need multiple independent domains. Then moving back images is much easier as it would just a matter of selecting the relevant images in the catalog and duplicate them back.

[RiaanBadenhorst]

Hello

The requirement you're describing is addressed by HA, not DR. The purpose of AIR is to make a copy on a remote site to protect you against a disaster in the main site. The scope of this disaster it is meant for is complete site failure, which means your entire site is gone and you would rebuild it in due time.

During the rebuild period your services would be failed over to the DR site and they would be backed up there until the main site is restored. Once restore you could have an SLP that replicates all backups that are being taken to the main site.

At some stage you would fail back.

The scenario you describe is best handled by having another appliance in the main site, or as mouse suggested, having one domain spread across sites. The catalog would be kept outside the appliance on a dedicated master server. 

[quebek]

Hey

Of course I will be using AIR to replicate the hot-catalog backups to DR site. So I could have used this backup to recover policies, SLPs, backup image lists... But how to replicate back these backup images... As of now I don't see any possibility to 'repopulate' the MSDP contents from DR site...

And about HA - well this is not HA requirement it is just a question how I can have these backups back on the source appliance... With NBU and tapes only it was only a matter of restoring the NBU catalog - and all media needed to perfrorm any recovery was available (if it was properly offsited). No need to figure out how to put back these images into MSDP. I am finding this as a gap in current NBU. Also I don't want to have any tapes on such remote site... So ... any other ideas how to have backups made at source site back to newly reinstalled NBU appliance?

[RiaanBadenhorst]

Hi,

I understand, but think of it this way. If you were using tape and offsiting a copy as well, then if your main site tapes get destroyed, you'll need to go to DR site and duplicate all those media again and take them back to main site. This duplication process in an AIR scenario is handle with nbreplicate as others mentioned (and its a bit tricky).

That is however the state of things. My suggestion for HA would have two appliances at main site which protects against appliance failure, but not site failure. 

[quebek]

OK got the message... So the POST DR activities are tricky to have it done... the DR will work but how to find out ourselfs in post DR situation. There will be times that some one will request the restore from backups taken to original NBU appliance - and I can't imagine to reach out to the DR site in evey single time - espiecailly for lager restores where the WAN link will be biggest obstacle.

[RiaanBadenhorst]

Hi

Exactly, but that is one really needs to look at the purpose of the DR site is. You should be using DR, when there is nothing left at PROD. That is it true use, but due to costs constraints people adopt a hybrid type of approach. 

Assume we were not talking about your backup appliance but your storage array in PROD. If it was destroyed you'd be in the same scenario assuming you had one in DR with all data replicated to it. You'd need to get all the data back to PROD, lun by lun.