Import Time

[jounix]

Recently I've done some import testing for DRP. Apon one test I import 6 tapes, with about 150gb, and it took roughly 1 1/2 hours to complete. The second test I did was with 3 tapes, 375gb, and it took almost 3 hours to import.  (These are just import times/not including the restore time.)

I though the import time would have decreased with the fewer number of tapes. We are actually attemtpting to move TB's in the future and import times this long would take forever.

Do you know what factors can determine the import time? Is it based off of number of tapes or data itself? What, if anything, can be done to get faster results?

[RiaanBadenhorst]

Why are you importing all your tapes?

You can simply restore your catalog backup at your DR site, then you have all the catalog information about each and every tape/backup image from production. You dont need to import each one!

You should only ever import a tape if it has expired (the retention period has run out). Say for instance you have a backup with one week retention period. You send it offsite (write-protected), and then someone says to you in two week they actually need the information on that tape. Because the retention on it was 1 week, NetBackup would have expired the images in the catalog after one week, and flagged the tape as available. You then need to import, phase 1 and phase 2 the tape. After that you can restore the images.

To answer your question about the time, it takes long because the data on the tape has to be read and placed in the catalog.

I trust that is clear now.

[Max_Booth]

Definitely agree.

Importing all your tapes will take a seriously long time.

In answer to your question - it's the size of the backups (amount of data on tape) which is the main factor of the import phase. During phase 2 of the import it has to recreate the files-files in the image catalog which requires reading through the entire backup image to create (since it reads through to catalog the contents of the tape - block positions, files, metadata etc).

max.

[J_H_Is_gone · Accepted solution]

I have see a difference in imports as well.

For me I have seen tapes with a lot of little files take longer has it has to write info for each file into the db/images.

If you have a working DR site and don't want to import tapes to it you could try this.

http://exftpp.symantec.com/pub/support/products/NetBackup_Enterprise_Server/263423.pdf

Which is doing just part of the catalog (the db/images dir)

I have used this and I no longer have to import at my working DR site.

I went one step further ( I have aix masters) I have an RDIST setup to everyday send the db/images dir from my prod master to another dir on my working DR master.

Then when I want to do a restore I just make a link

/usr/openv/netbackup/db/images/servername -> /prod/db/images/servername  (where prod/db/images is the dir where RDIST put the file)

the image file contains all the info of what tape the backup is on.

You just write protect your prod tape, put it in the robot and put in a separate vpool (to make sure your DR server does not use it)  and then when you start the bar and use your prod server as the source it follows the link and sees all the backups and you can do the restore.  Works great, and cut out hours of me having to do imports for restores at my DR site.

edit:  left the servername off of the link.  makes a difference.

[jounix]

Its not technically a DR site... I'm taking a database from one data center and then taking those tapes to another data center and importing them into that data centers current catalog.

We are basically trying to build another server at a different location and need a way to get that data to that site and restore it there. These databases are roughly 550gb - 1tb. So, time is of the essence.