Streaming Question

[Tommojunior]

Hi Folks,

I'm coming up against a challenge with the definition of streams in a policy. A trivial example of the situation I have is as follows:

Say I want to backup the contents of /home on a unix server (NB4.5MP3) HPUX. I've got 5 drives which I want to use and 4 of my users have large amounts of data, but the others do not:

/home/mark - 50 gig
/home/peter - 50 gig
/home/jane - 50 gig
/home/paul - 50 gig
/home/ruth - 1 Meg
/home/john - 1 Meg
/home/matt - 1 Meg
/home/cath - 1 Meg

All these home directories are within the same filesystem (/home) so Auto-Streaming won't help me. I set about writing my file list in the policy to try and make the most of my available drives:

NEW_STREAM
/home/mark - 50 gig
NEW_STREAM
/home/peter - 50 gig
NEW_STREAM
/home/jane - 50 gig
NEW_STREAM
/home/paul - 50 gig
NEW_STREAM
/home/ruth - 1 Meg
/home/john - 1 Meg
/home/matt - 1 Meg
/home/cath - 1 Meg

This will work fine for me, but what if I get 100 more users (all with 1 Meg each), I don't want to specify each of them explicitly under the last stream, what I want to have is:

NEW_STREAM
/home/mark - 50 gig
NEW_STREAM
/home/peter - 50 gig
NEW_STREAM
/home/jane - 50 gig
NEW_STREAM
/home/paul - 50 gig
NEW_STREAM
/home - All the other users

However - This doesn't have the desired effect, since what netbackup does with the last stream is to back up /home/mark and friends a second time. See what I'm getting at here folks, is there any clever way of saying for the 4th stream:

NEW_STREAM
"get me stuff in /home that hasn't already been captured by the other streams"

Thoughts on a post-card please!

ooh - and just in case your thinking about platter hammering etc, the real senario I have is on an EMC rig, so I'm not that fussed about giving it a hammering.

Cheers,

Mark Thompson

 

[comtec17]

If this is on an EMC why do you care on the streams? Is your host that slow to backup?

I would suggest the following:
Create 1 class for these:
NEW_STREAM
/home/mark - 50 gig
NEW_STREAM
/home/peter - 50 gig
NEW_STREAM
/home/jane - 50 gig
NEW_STREAM
/home/paul - 50 gig

Then create a different class and make an exclude list with the following entries: /home/mark, /home/peter, /home/jane, /home/paul
then
NEW_STREAM
/home/*

 

[PGPhantom]

Bah-Humbug - Comtec17, you beat me to the draw

 

[Tommojunior]

Cheers Guys!

Am I missing something very obvious here? Is there a way of spanning a single filesystem backup across more than one drive without breaking it down into streams. Auto-streaming won't do this will it?

comtec17 - what do you mean by seperate class? do you mean a seperate policy?

Ta,

Mark

 

[Tommojunior]

The opposite of multistreaming I guess is what I'm after. Not multiple policies going to a single drive, but a single policy going to multiple drives.

I can't easily split the file-list across more than one policy, because I'm using pre and post script (I'd have to write a clever bit of scripting for one job to see if the others had started).

 

[comtec17]

I was meaning creating a different policy. If you are trying to capture what is left in home then either list all of them in a stream or create a seperate policy and create an exclude list to exclude the other filesystems. You will not be able to capture the "other stuff" in /home without it.

 

[Tommojunior]

So no way of putting a single filesystem to multiple drives without breaking it down into streams?

I can see why NB works like it does, its kind of to protect users against hammering a disk jumping from platter to platter, but the problem really isn't there when you are talking about SAN storage.

I guess what I'm talking about is kind of multiplexing in reverse (one big pipe into 5 small ones as opposed to 5 small/slow clients to a single big pipe). I guess my data would be scattered across 5 pieces of media (increasing risk).

Am I way off the track here?

 

[comtec17]

You can put a single filesystem to multiple drives by using the ALL_LOCAL_DRIVES directive. Each filesystem will use a drive.

 

[Tommojunior]

(Working late tonight aren't we!)

Agreed - each filesystem will use a drive, but if I'm only backing up one filesystem (/home in the example) that doesn't help me.

Ofcourse I could set about splitting the lvol into pieces and re-designing the LUN layout on the rig, but this is a huge parcel of work with significant risk in itself.

Hmmm...

 

[Tommojunior]

So, lets simplify the example, caus it sounds like I'm missing something obvious:

I have one filesystem (/home) on an emc rig which I want to backup to 4 drives. Can I configure NB for multiple concurrent streams without using file list directives?

 

[comtec17]

No.

 

[picapiedra]

Hi guys! One thing about what comtec17 said (sorry about my lack of experience):

"You can put a single filesystem to multiple drives by using the ALL_LOCAL_DRIVES directive. Each filesystem will use a drive".

I have one policy with the SYSTEM_STATE and ALL_LOCAL_DRIVES directives. W2003 client, not multiplexing, it has C:, D: and E: drives, E: is on a Hitachi SAN, and it runs in just one drive (150Gb) taking up to 7 hours to finish, I´m just thinking about multistreaming E:\ drive in the File List tab, but at the moment it just use ONE drive.

What´s wrong? Should it be using more than one just because the ALL_LOCAL_DRIVES directive? I think I´m missing something.

Thanks in advance.

 

[picapiedra]

Wait!
Probably because I haven´t checked the "Allow multiple data streams" box in the Attributes Tab?

¿¿!!??!!

I´m sorry... and always learning...

Thanks anyway, guys.