Video Screencast Help

NDMP Policy best practice

Created: 31 Dec 2013 • Updated: 13 Mar 2014 | 6 comments
This issue has been solved. See solution.

Hello,

While I have some experience with NB, I really haven't done much in terms of NDMP backups, so I have a question...

I just started working in a 7.5 environment (RHEL on the master and media servers). The person who initially setup the policies is no longer around.

The standard policies are what one would expect...one policy encompasses all Linux clients that have the same backup needs in terms of schedule, backup selection, etc... Same for Windows and other policy types. However for NDMP, there is a policy for each subdirectory under /vol. So there are a lot of NDMP polices with the same attributes, schedules, client, but the only distinction is the Backup selection. Since I do not have much experience with NDMP, I'm wondering if this is typical. Why not allow multiple data streams, and use NEW_STREAM directive, and specify the various /vol/dirs in one policy?

We are doing remote NDMP, backing up to a Data Domain, using OST, via a media server. (and SLPs to replicate)

 

Thanks

Operating Systems:

Comments 6 CommentsJump to latest comment

SymTerry's picture

Hello,

You would be correct. I would test it, but you can try multi streams and see if that works for you.

SOLUTION
Will Restore's picture

We do something similar.  A few of the smaller directories are grouped in each Stream however rather than each directory individually.

Will Restore -- where there is a Will there is a way

watsons's picture

Yes, what you think makes complete sense and I would do the same.

As of why the person before you configured that way, it's possible they are trying to track down an issue with backup taking too long to complete or ended with an error. With all using the same policy, even though with multiple streams, your parent job would be a status 1 rather than 0. I am not particular about that but I have seen people who are - they need a definite status 0 (probaly for reporting purpose).

LT2013's picture

Thanks all for the confirmation. I know with NAS backups they tend to be large so breaking them up is a good idea, but I figured using one policy with multiple streams would make sense.

rk1074's picture

Hi LT2013

You are absolutely correct!!!

We are also having a backup of a Netapp filer via NDMP where we are backing around 150+ volumes in a single policy. What we have done is that we have allowed multistreaming and clubbed  number of volumes into a single stream via NEW_STREAM.  in this way it lowers the number of child jobs being generated (150 to be precise in case if you do not use NEW STREAM option) for individual volume.

Andy Welburn's picture

There *may* have been valid reasons for separate policies initially (different schedules, different volume pools etc), but requirements do change over time so it may be that these are no longer valid & therefore you are quite probably correct in your assumption that they can all be grouped together within one policy.

 

When we utilised NDMP backups we had a mixture - small volumes grouped & larger volumes within their own policies. For us, this simply enabled greater control over exactly when a job could/would start. If a specific job needed to be re-run then it was also easier to do if there were not any other streams to consider (altho' a one-off policy could be created to cater for this)