Mounted filesystem in VMware backup

Hi Ramnagalla,

 

Thanks for your response.

 

In filesystem backup we have option called "Follow NFS" to backup NFS share. Is there any similar options in VMware based backup?

Nope.. does not have such options in Vmware backup type.. it can only take the local FS ,(not RMDs as well)

and it not because of netbackup limitation.. it is the limitation from Vmware..

Vmware can not take the snapshot of the non-local file systesm, also RDMs as well..

I know several/many of us are quick to defend NetBackup and quite clearly draw a line between NetBackup and issues outside of NetBackup...

...but sometimes it strikes me that perhaps sometimes we also need to defend (a little bit at least) other infrastructure partners/vendors too...

...so, I think, in situations like this... even we can't say that the inability to snapshot a RDM is a VMware limitation - because, IMO, it's not so much a limitation... and more simply to do with the functional capabilities of LUNs and block level storage in general.

@Kumarantg... my response is that straight/vanilla/plain block level storage simply does not have the functionality to snapshot.  The intregration of the entire IO stack, from application -> database -> advanced IO performance for tables and rows -> OS read/write APIs -> file system manager-> volume manager -> partition manager -> LUN -> SCSI -> SAS/FC/iSCSI/NAS/NFS/SMB -> OS driver -> HBA/NIC -> storage array -> parity-group -> RAID algorithms -> disk bus -> spindles (...and let's not forget the other software stack rising on the other side, on the storage array side, to manage the storage array view of pretty much the same set of tiers...)... well... to achieve a consistent point in time and safe and clean and flushed and quiesced (briefly) IO stack all before a snapshot can be  established (assuming even that a storage array chassis is advanced enough to incorporate a snapshot technology of any sort)... is actually quite difficult to achieve, even for one vendor, let alone several vendors in co-operation.  In short, it's quite difficult to reliably and consistently achieve that level of control, and so requires lots of co-operation and lots of development, and lots and lots of testing... i.e. big R&D budgets.  Customers with deep pockets not only want such functionality, but need it... their businesses are simply too valuable to not have it.  Not many need it, not many can afford it.  And until demand increases, then the price isn't going to drop.

And it gets even more complicated when we introduce multi-pathing, multi-fabrics, HA servers, clustering, active/active nodes e.g. (Oracle RAC or MS Hyper-V CSV) vs active/passive (simple MSCS or VCS), and then the other forms of active/active storage vs. active/passive storage... it can all get a bit tricky.  Everyone always wants to snapshot RDMs, and the like, hopefully now it's a bit clearer as to why that simply is not possible for simple/basic/cheap configurations.