Duplication to tape : how can I restrict total number of drives all duplication jobs use

[briggsy]

Hi,

We have an issue on one of our environments where we want to restrict the max total number of drives that any duplications from any media server can use is (ie. no of drives available -1 )

Environment

All servers are Windows 2008 R2, Master and media server 1 are NetBackup 7.6.0.4, media servers 2 & 3 are 7.5.0.6 (to be upgraded shortly)

1 x Master / Media server , 26TB san provisioned MSDP disk pool

3 x Media servers, each with 15TB san provisioned MSDP disk pool

1 Quantum library with 6 LTO5 tape drives, available to all servers via FC

There are 4 tape storage units, 1 per media server, each with max concurrent drives = 6, all are set to multiplex up to 32 streams/drive

There is a tape storage group, containing all 4 tape storage units

Problem

Initially (about 2 years ago) all backup policies went to SLP , so backup to MSDP then Duplicate to tape, various policies pointing to 1 of the 4 media servers, duplications to the tape storage group.

As data volumes grew we had issues with disk pools filling, so now we put some of the larger SQL backups directly to tape, leaving the MS-Windows, Unix and VMWare backups to use the SLPs.

Problem we have now is that we duplicate so much data that the support guys are constantly having to abandon duplications to allow tape backups to grab a tape drive, as all 6 drives are nearly contantly in use.. Ideally we need more tape drives but getting the customer to pay is another matter completely !

So what I'd like to do is limit the total number of drives available to duplications to 5, therefore leaving 1 drive always available to backups, unfortunately I see no way of restricting a tape storage group to 5 drives.

I can create additional tape storage units with a max of 5 concurrent drives, but that just means each individual media server can only duplicate to 5 drives, so with 4 media servers all drives will still be used.

I could also create additional tape storage units with a max concurrent drives of 1 , but that'll mean only 4 tape drives being used for tape duplications, and a max of 1 per media server....again not what is required..

Any advice appreciated !

Also, on the SLP parameters (now available on the console - thanks for that !) ...

Tape (and Disk) resource multiplier values are set to 2  - what do these settings do ?

Many thanks

[Yasuhisa_Ishika]

No way to reserve tape drive to specific opetation. Setting higher pripority in SQL policy is enough for your requirement?  SQL job will start immediately afrer one SLP deplication job ends.

[briggsy]

Hi, Thanks but changing the priority won't help, problem is that quite often all 6 drives are in use for duplications when the backup starts, so the support guys quite often have to abandon duplications to free up a drive.

[Marianne · Accepted solution]

The issue that I see here is that you have insufficient resources to ensure duplication jobs are completed before the new backup window starts.

The solution here would be to have a total re-evaluation of your backup infrastructure.
Neither disk volumes nor the tape drives can cope... 

One problem I see with tape STU config is that all of the media servers have the potential to use all 6 drives concurrently, leaving the rest of the media servers with nothing to do.
This could be part of the problem - one media server trying to use too many drives simultaneously, resulting in terribly slow throughput and resulting backlogs.
Change the max number of drives to 2 in each STU.
All of the media servers will still be able to use all 6 drives, just not all at the same time.
Hardly any media server that I have seen over the years have the ability to properly stream more than 2 drives simultaneously. LTO5 drives should be streamed at about 100MB/sec.

You should also look at SLP config carefully to ensure duplication from disk to tape happens locally on the relevant media server - no network duplication from media1 to media2.

In NBU 7.6, you can configure duplication windows that can suspend duplications at the end of the window, but that would create a further backlog.

Another workaround would be to dedicate a single media server for the large SQL jobs and zone one tape drive for this purpose. 
Either one of the existing media servers or else turn large SQL clients into SAN media servers, sharing 1 or 2 dedicated tape drives.

PS:
You may want to look at TAPE_RESOURCE_MULTIPLIER setting in LIFECYCLE_PARAMETERS.
See: https://www-secure.symantec.com/connect/forums/slp-using-just-one-drive