Restore only changes & new files without overwriting?

[Calzor_Suzay]

We quite often have to provide data to 3rd party companies that trawl through data for key words for legal court cases. They charge on a per GB basis so the least amount of data we provide them with the better.

Is there a way within NBU 6.5, 3rd party or via another method to restore only changed (Renaming) or new data without overwriting?
For example we restore Jan '10 month end in full from a full backup. We then only need to add the changes to files and new files on the following months backups say Feb '10, Mar '10 which are full backups.

In some cases the month end restore could be several hundred GB so reducing this could make a big difference.

[Andy_Welburn]

in the backup policy(ies) specifically for that reason I'm not aware of any procedure/software that would make it possible.

e.g. Policy_One Schedules:

Jan FULL (month-end)

[[... differential incrementals throughout month if required.]]

Feb CUMULATIVE (month-end) i.e. changes since last FULL

Feb FULL (month-end) baseline for next 'set' of incrementals

This would obviously increase your environments workload & there is also the possibility that some changes could be missed between the cumulative & full.

[Calzor_Suzay]

Yeah I had looked at this but would cause a big increase in workload, tapes etc.

It also doesn't cover historical backups which can be as far back as 10 years.

I think it's one of those things I know can't really be done but need to ask

[Andy_Welburn]

back as 10 years."

Ah! So you've just been given this job & need to look at historical data & not just for backups carried out from this point on......

Might just have to bite the bullet, altho' there may be more ideas out there!

[Calzor_Suzay]

Not just been given it but it crops up quite often

[J_H_Is_gone]

I use to do this way back in the 80's with 1650 mag tape ( yes I have been around a while)

When needing to do something like this you restore in reverse order.

So you need

Jan month-end

Feb month-end

March month-end

But want the most recent copy of all files

--------------

Restore March month-end first - this is because it is the most recent copy of all files out there.  So now you are only missing files from Jan and Feb that have been deleted.

So now you restore Feb month-end - NO overwrite - it will only restore files that had been removed by the time the March backup ran.  It will not overwrite any March files ( as they are the newer ones)

So now you restore Jan month-end - NO overwrite - it will only restore files that had been removed by the time the Feb backup ran.  It will not overwrite any Feb or March files.

So now you have all files, and the most recent copy of all files.

Does that fit in with what you need to do?

------

now as to getting all copies of files that have changed - that I cannot help with.

[Will_Restore]

I like your "backwards" thinking

[J_H_Is_gone]

If you only have those 3 monthend tapes avail.  When you choose all 3 in the BAR you get just that, the most recent copy of all files.

The only time it does not work with NB is if you don't want to do the incremental in-between as you cannot no 'un-select' them.

------

In thinking about that, it also comes to the option of

"Restore with a temporary file name"

But with that it would still restore everything only you would have 3 copies of it.

[Calzor_Suzay]

I'm not sure how different this is from restoring Jan in full then Feb over writing it all then Mar over writing it all, you'll still end up with the latest file and restore won't remove files from Jan, Feb that aren't there in Mar but anyway no it doesn't help I'm afraid

What about a file that was there in Jan with x data then it's there in Feb with xx date now it's there in Mar with xxx I need the three different versions of the file with x xx & xxx, people try and cover their tracks

[J_H_Is_gone]

Your way you are restoring all files, for each month.

If you choose all 3 months at once, or do the backward restore you are restoring all files for March, but only the missing files for Feb and Jan (fewer files)

You issue is still wanting all copies of files that have changed.  In Netbackup there is no way to only ask for files that have changed.

You can however still get all three versions on disk at once.  It would then be by hand to figure out if they had changed.

If you do the restore from Jan all files.

Then do the Restore from Feb all files, but choose to restore with temporary file name.

Then it will restore any files that are already on disk, with the same file name PLUS a random number append to the file name to make it unique.

then do the same with March backup, all files restore, with temporary file name.

The only thing here is that you get all files from all 3 months, not just the ones that have changed.

[Calzor_Suzay]

I see what you're saying about fewew files now.

I've tried restoring in full Jan then Feb the Mar etc and trying to compare them with 3rd party software but with the amount of data in question most of the time they fall over in a big heap.

I guess the possibility exists to restore everything in full to different folders then use Robocopy to do a log mirror comparison of March to Feb, us the log to script and move the files elsewhere, delete Mar then do the same with Feb to Jan, very long winded and painful :(

[Will_Restore]

Judy's "backwards" solution is quite elegant & should do the trick.

[Will_Restore]

what I meant was

it works out of the box

and your thinking is out of the box

[J_H_Is_gone]

When in the BAR for each file it list the backup date and the modified date.

So what you need is to restore the Jan then in the Feb backup find all the files that have a modified date grater then the Jan backup ( say 1/31/2010)

I know the windows BAR search is a crap shoot trying to get it to work.

I have not used the Java console to know if it would work.

But I know there is a command to list all files in a backup image.

So if you could find that command, then find a way to list only files with a greater modified time.  On UNIX I would use awk.  And put that into a file. 

Then you would want to removed any lines in the file that are just directories. You don't want to restore those because it would restore everything in the dir.

Which leaves you with files that have changed.

Then restore those with a temporary file name.

So you would end up with

filea.txt

filea.123456.txt

then do the same for march.

I don't know cli that well, but I know there is a command.

So maybe change your thread name or start a new thread with Restore - need cli help.

If you can just find a way to find the files in the backup image with a modified time greater then the previous backup you would have a good list to restore.  That list can be used a cli to do the restore.

[alazanowski]

as an alternate to all of this, you might consider using an e-discovery solution. Enterprise Vault does very well on searching through archives for files with specific keywords in them and makes it export friendly for legal cases. Just a thought; seeing as doing this manually with netbackup and then paying someone else to go from there is also quite annoying.