Enabling NDMP backup of storage (NetApp) from NW

The standard Networker Administration Guide (part of the downloadable documentation suite pdf) is fairly comprehensive and has a chapter on this. I believe it is a separately licenced option and you need to confirm your architecture, are you going to backup to SAN/SCSI NetApp connected NDMP drives or via IP to standard Networker drives on your backup server or storage node. There are pros and cons to each, directly attached or DDS shared NDMP drives on the NetApp run quickly and put no additional load on the IP network, however you cannot run multiple data streams to a single NDMP drive so volumes backup in a serial fashion unless you allocate a drive per stream.

Indexes and bootstrap records cannot be written to an NDMP drive so you need to schedule the index backups to a separate tape pool if backing up directly from the NAS filer

NDMP tape drives and media pools need to be defined as such when configuring the robot / VTL and you should set the drive parallelism to 1 (see above).

There are some additional parameters when defining NDMP clients these depend upon the configuration, but generally for NetApp are:

remote user / password

backup command: nsrndmp_save -T dump [add -M if 3 way to Networker server]

tick the NDMP box

Application information:

HIST=Y                              - Maintain indexes
UPDATE=Y                        - Maintain dumptimes - allows level x backups
DIRECT=Y                         - Allow DAR tape file positioning
EXTRACT_ACL=Y               - Record ACL information

Finally NDMP (ufsdumps basically) do not support incremental backups but you can achieve the same effect by using a scheule that does:

Full, level 1, level 2, level 3 .etc.. Full

Good Luck

Thanks for the detailed info.

Let me ask you a very basic question:

We currently have a backup server running NW 7.5 and several clients are backed up by this server to a tape library.

NAS volumes are currently backed up,  by mounting  the volumes  on to a server and the server is then backed up from NW.

With NDMP (license),  I  understands we can avoid the middle server and the volume can be backed up straight  by the NW, right?

What do we gain from this?   Do I need to set aside a separate drive for this?  Or the NAS will be just like any other clinet server for NW?

If you look at the EMC supplied documentation (Networker Server Support for NBMP in the Administration Guide) you will see pictorial representations of a number of permitted NDMP configurations.

NDMP is simply a protocol that instructs the NetApp to initiate the backup process, the NetApp then perfroms a special snapshot of the volume(s) in question, .backup_snapshot, and backs thise up using the *NIX ufsdump command. This dump can be taken either:

1]     To a directly attached (NDMP) drive attached to the NetApp - this can be a dedicated tape drive or shared using Networked Dynamic Drive Sharing (DDS)

2]     Back up to the Networker backup server

3]     Back up to a Networker storage node

The latter 2 examples are termed DSA backups and package the NMPD (ufsdump) datastream into the Networker open tape format.

Option 1 supports tape only and since NDMP tapes are written in a different format to Networker Open Tape format require a special NDMP flag on the drive and a pool of NDMP tapes. NDMP dump can only single stream to each drive but can write at native SAN / SCSI speeds and sends only metadata over IP - possibly the fastest option for backup.

Opitions 2 and 3, allow the use of advanced Networker features like multiplexing, backup to disk, cloning etc. But all data is sent to the backup server / nominated storage node over IP - they can use existing Networker tape pools.

Any of these NDMP configurations mean you do not have to mount the NAS devices on an intermediate server to perfrom backups. Specifying 'All' in the NDMP client resource will dynamically discover any NetApp mountpoints and back them up.

The main reason for using NDMP is to speed up the backup process where there are a large number of small files to save. Whether you use a directly attached tape drive or a DSA backup to a Networker node depends on your IP network vs. SAN bandwidth.

In our case the elapsed backup time for a full backup of our corporate NFS share (5 million files / 270GB) using mounted filesystems over 1GB shared IP took around 23 hours, using NDMP to a directly attached SAN tape drives it now takes around 2 hours.

Very interseting. Thanks for all the details.