both yes

As far as I could find out you have to choice of NDMP backup the whole nested mount file system or the component file systems.
We recommend backing up the component file systems though - since if you backup the whole NMFS on a full restore you have to be careful to get the structure in place first (or it will most likely fail when running out of disk space)

In terms of checkpoints - just remember the caveat in the SnapSure manual:

Note: EMC recommends that you do not create checkpoints under nested mount points. If
you do so using Volume Shadow Copy (VSS), you cannot restore the PFS from a
checkpoint of the PFS without first permanently unmounting the checkpoints and the PFS
from the NMFS. Then you must mount the PFS and its checkpoint to another mount point,
perform a restore with VSS, then reverse the process to mount the PFS and its checkpoints
on the NMFS again.

This is applicable for a full restore (revert) - i.e. resetting the PFS to a previous checkpoint state and not for individual file restores

Rainer ..thank you. I have a question though ..Let's say i have a NMFS with 3 component file systems, how do i take a snapshot of these 3 file systems at the same time to have a consistent snapshot ? Correct me if i am wrong but right now i can't create checkpoints of multiple file systems at the same time. My concern is that if i have multiple file systems mounted underneath my root file system for example

\Fruits
           \componentfs1
           \componentfs2
           \componentfs3

is it possible to have a file where part of the file resides on componentfs1 and another part of the file resides on componentsfs2. In this case i would have to have a consistent snap of all component file systems?

Thanks again

you'd have to snap the component file systems sequentially.

but no - a file cannot span multiple component file systems.

a NMFS is like when on a Unix local file systems you mount them hierarchical with the expection that you can export the NMFS in one go - whereas normally NFS and CIFS exports cannot span file system boundaries.

So you can have more than 16TB under one export (and therefore one client mountpoint)

You'll see each component file sytems as a subdirectory - similar to a treequota.

hope that clears it up

i just created a test nested setup and like you said ..i do see component file systems as separate directories. I actually wish i did not see them as separate directories but rather as one big file system where on the back-end it's really two file systems. Let's say i need to create a huge directory for users home directories. I wanted to utilize NMFS where i could create a 2T CIFS share where in reality it would consist 4x500G component file systems, so that when i run NDMP backup i could use 4 tape drives. Since these 4 component file systems will show up as 4 unique directories it will not be possible to have all user home directories in one location, for example

\HOME
  \user1
  \user2

My windows admins share HOME directory, they do not share each users home directory. So HOME is shared and when user connects to \\servername\HOME ..all they see is their home directory (we enabled access based enumeration). Not being able to put all user directories under one directory is not going to work. Any other ideas/suggestions ?

Thanks

why not let the Celerra do the work and use the builtin Home Directory functionality ?

See the "EMC Home Directory White Paper" on Powerlink.

Home > Support > Technical Documentation and Advisories > White Papers > All White Papers

This way you can put users in as many file systems you like, you could use different speed fs and even move users from one fs to another without having to change the Windows AD homedir setting

i've played with HOME directory functionality on NetAPP ..which is the same as on Celerra. In my case i may have file systems that will not contain home directories, do i have any other options. My ultimate goal is to be able to backup 2T file system with 4 tape drives ..concurrently. So i was hoping using nested file system would allow me keep it transparent from user perspective (they just see one big share), while i can back it up component file systems in parallel.

Thanks

lets say similar - there are some differences

besides NMFS the other way I can think of is using symlinks or even widelinks with DFS root to build your own "namespace"

Just try it. Just be aware that once you enable symlink following a NFS client with the appropriate access and knowledge of the file systems structure could create a symlink pointing to a directory that you didnt want to be shared.

very fragmented environment here ...many people's computers are not even joined to AD, so i am going to stay away from DFS and symlink can get too messy. So let me ask this question, it's very nice that Celerra can provide these huge mult-terabyte file systems, but how are we supposed to back them up ? I run full backups every two weeks and rely on snapsure for my "incremental". I have dedicated tape SAN with Celerra connected to a CDL, yet backups still taking forever to complete. I know some backup applications offer synthetic full but i am not the proud owner of one of those applications

we are going to eval Avamar, any experience with Avamar and Celerra that you can comment ?

Thanks

I think the larger systems get the less often full backups are taken

Avamar is a good example - its like never doing a full backup
plus the deduplication so that you can afford to have more version available

of course the question is why you take full backups - if it is in order to do a full restore after a complete data loss then these days its more common to just replicate everything to a second system.

exactly ..i do my full backups for complete restore in case file systems gets corrupted ..etc, i am hoping Avamar will show some promise in this area because these full backups are killing me.

NDMP and NMFS don't play well together:
I know this is an old thread, but thought I'd share my experience with NDMP backups and NMFS. Our NDMP backups with CommVault 7.x are working fine with everything (CIFS and NFS, and file-based and volume-based), except for Nested Mount File Systems. I get the error "not backupable object" on the parent NMFS file system. Then the entire job fails.

Oct 4 12:43:03 2009 CS_PLATFORM:NASDB:ERROR:11 Command failed. Command: /nas/bin/fs_ckpt id=145 -name automaticTempNDMPCkpt46-145-1254674574 -Create -option automount=yes,mountpoint=/automaticNDMPCkpts/automaticTempNDMPCkpt46-145-1254674574 Error: Error 2236: fs_nmfs_ove_wholesale : not backupable object

We then get subsequent errors stating that the automatic NDMP checkpoint cannot be deleted.

Regards,
Ted Pound
Atlanta, GA

did you configure CommVault to backup the component file systems and NOT the NMFS root ?

what DART version are you using ?

And how do I backup the component file systems ?

I just try with the Netbackup .. it does not see component filesystem mount ...

12/17/2009 9:23:25 AM - Error ndmpagent(pid=13208) NDMP_LOG_ERROR 0 Invalid backup path        

Yes, we were just attempting to backup the component file systems and not the parent.

Ultimately this issue caused memory leaks for the Celerra and we ended up having to reboot due to it. Not good in 24/7 production environment.

Hi

Though old post, but have same questions...did anyone get NMFS work with NDMP backup?

Looking at splitting the cifs and use nmfs to extend number of concurrent ndmp backup...

thanks folks