Active IQ Unified Manager Discussions

DFPM: How to avoid that a new Dataset is creating two relation instead one? Non Qtree "/vol/nas_daten/-" which is superfluous !

hondasuschi
3,524 Views

Hi may somebody can help

When creating a new Dataset and attaching backup policy like qsm, then a new snapmirror / vault is with two relations-ship instead only one is created.

-One relation-ship with qsm source  "node1:/vol/HR_Data/test          node2:/vol/HR_Data/test"

This relation-ship is the only wished . We do not want to have the relation"/-" means all non qtree data! Never non-qtree data are on normal Systems existing on root

volumes like /vol/volz/manualDir. I think the intentions of development, were to avoid that some non qtree data is not saved. But in my opinion this is a bad design,
as no normal Netapp Admin will ever create manual Directories (non Qtree Dirs) on a root volume.

Here some example:

node1:/vol/HR_Data/-             node2:/vol/HR_Data/HR_Data_node1_HR_Data        Source         00:01:54   Idle (unwanted realtion but created by new dataset )
node1:/vol/HR_Data/test          node2:/vol/HR_Data/test                         Source         00:01:55   Idle (only this qtree contains data)

The "/-" means that all non qtree data will be saved. But qtree data are not saved at all with that. (This will guarantee that manual created Directories/Files will be

saved, but this never happens in practice therefore why ?)

To overcome this, a second relation with the effective data qtree will be created. As consequence we get always two relation instead one.

Of curse workaround would be to create a manual volume & qtree an then selection while creation of dataset (not provisioning) manuall qtree as source.  Then "/-" would

not be created!


How we can change this wrong behavior of creating this non qtree "/-" relation which is completely wrong ? Is there an option available ?

5 REPLIES 5

reide
3,524 Views

Juerg,

In the dataset's physical members, are you specifing the parent volume (i.e. node1:/vol/HR_DATA) or are you specifying the exact qtree you want (i.e. node1:/vol/HR_DATA/test)?   If you specify the exact qtree as the physical member, it should only create a single relationship.  It would not create a relationship for /vol/HR_DATA/- because it wasn't specified as a physical member in the dataset.

If you specify the parent volume as the physical member, it will create a relationship for any/all qtrees that live within that volume.

Reid

hondasuschi
3,524 Views

Hi Reid  Is for a new Data Set wizard only. Of curse if you select an existing qtree then the "/-" is not created!  But if you take the dmpm (blue java gui) and select add new data set wizard , then this dataset is creating a qtree but the snapmirror relation is not only build to this only secifc qtree what is liked.   Unfort. a second non disired "/-" non qtree realation is created as well. This is abolutelly not disired! It doubles the number of snapmirror for nohing.

kryan
3,524 Views

Jureg,

OnCommand UM Core (DFM) does not support SnapVault (or QSM in this case) of volumes without adding the "/-".

For example, it is not possible to import a volume SnapVault relationship into a dataset if it was created as "node1:/vol/HR_Data" as it is not recognized as a valid relationship although it is certainly possible to create within ONTAP.  The reason for this behavior is that any restores of a qtree protected under such a relationship is not a qtree upon restore, but merely a folder.  Therefore DFM will not create this style of relationship as it would not be possible to restore a qtree as a qtree.  Instead, when protecting a volume using a "Backup" style protection policy, it will create a relationship for each qtree within the volume in addition to the "/-" volume qtree style relationship.

Thanks,

Kevin

hondasuschi
3,524 Views

Hi Kevin  Still not the solution. I perfectelly understand that qsm and qsm do not support node1:/vol/volx as a snapvault or snapmirror source in dfpm. But would be possible in commandline. The why you expained right.  Is not this question here. If you are going to create with dfpm wizard a new dataset the at first is going to create a new qtree and never creates a volume as source exp. >new Dataset creats >/vol/examplevol/exmple_tree_datsetname  The only only problem is why beside of creating a new qtree as source which is right! Why beside of that, we get a non qtree saving relation which is absolutelly not required? "This is a second snap-mirror/vault whith /vol/examplevol/- means all non qtree dirs will be saved. But you never have non qtree data on root volume, therefore the question why we get this non qtree "/-". We only have a new dataset. A new dataset is always creating a qtree never a volume saving relation, thus this "/-" is superfluous.

kryan
3,524 Views

I misunderstood the use case - I thought you were attempting to protect the volume containing a qtree, where what you observed would be expected.  Instead it seems you are only adding a single qtree to a dataset for protection but ending up with both a qtree relationship in addition to the "volume/-" relationship.

Public