Question regarding backup of NetApp Filer with Backup Exec 2012 to Dell ML6030 tape library over FC

[G4v1n]

HI,

Looking for clarification on the following.  We have been informed that we need the SAN storage options in order for a NetApp filer and Backup Exec server to be able to share a robotic library, to be able to backup via FC.

We have a NetApp FAS3220 filer running 8.2P5 7-Mode.  This is our SnapVault target which we want to backup to the Dell ML6030 tape library so tapes can be removed for offsite storage.

The backup server is a Dell PowerEdge R620 (Win2k8 R2 SP1) connected via fibre to 2x Cisco Nexus 5548.  The Filer has 2x FC (ports 0c and 0d) connections with one to each nexus 5548 switch.  There are 6 tape drives in the library connected by FC to the Nexus 5548's.  Drives 1, 3 and 5 to one nexus and drives 2, 4 and 6 to the other 5548.

Zoning is configured on the 5548's and within the OS tape library/drives are correctly discovered and identified.  Backup exec also correclty picks up the tape library/drives.

If I configure the filer as an NDMP server via the Configure Storage - Network Storage - NDMP Storage wizard.  The Robotic library and the 6 tape drives then also appear under the NDMP device in the storage tab so it looks like there are 2 libraries listed.

When I attempt to do a test backup from the filer I am successfully able to see the volumes, select a volume and pass credential check.  The job is successfully created.  However when the job runs the status remains at the Queued stage on the Backup and Restore tab, and on the storage tab the tape drives never seem to load a tape.  All that happens is that backup exec seems to be cycling through the available tape drives with the status changing from Online to Active to Online.

At this time the ndmp status on the filer shows the following

blk-ntap3220-01a> ndmpd status
ndmpd ON.
Session: 30375
  Active
  version:        4
  Operating on behalf of primary host.

  tape.device:    nrst0a

  mover state:    Idle
  data state:     Idle
  data operation: None

So it looks like nothing is happening.  This is what we have been told we need the SSO for as currently the backup server and the filer cannot determine who has ownership of the tape library.

Is this the case?

Regards

[Colin_Weaver]

a) If there is only one media server then SSO is not needed

b) Is the robotic arm part of the library zoned to the netapp(s) and the media server or just one of these. If it is not attached to all of these then jobs will go queued. Basically control of the arm is not centralized to the media server, instead the job distribution is centralized but each host (Netapp or media server) actually takes over job processing and the host must therefore be able to control the robotic library arm directly as otherwise we will not be able to move a tape to the required drive. Typically I think we often recommend that it is better to zone all drives and the arm to all hosts that may need to run jobs and then let BE do the management as to when the devices are available (although it is possible that you can still zone the drives to indivdiual hosts.)

c) if the library has 2 tape drives you will probably need a Library Expnasion Option License (LEO)

d) there are some limitations with regards ONTap 8.2 and 7-mode/c-node setups in that CDOT is not supported and we therefore only support basic clustering (node scoped ndmp) . Please note I am not sure of the exact meaning or relationship of the terms I state in this point I am just repeating information that came from our QA testing teams with regards new functionality in OnTap 8.x

[G4v1n]

Hi Colin, 

Thanks for the response.  Reagrding the points you make a few clarifications from my side

a) By media server do you mean the backup Exec 2012 server?  If so then yes there is only 1 media server.  However does the NetApp filer which is the SnapVault Target, and the source for the backups to tape qualify as a media server? The reason I ask is when I added the filer on the storage tab it now shows with the robotic library dispalyed under the filer and then again with the Robotic library displayed seperately.  

b) What do you mean by "Is the robotic arm of the library zoned" ? The library has 6 tape drives.  Each with a fibre cable into the nexus 5548 switches (3 drives in one switch and 3 in the other switch).  The library was cabled as per page 14 of the Dell ML6000 getting started guide

ftp://ftp.dell.com/Manuals/all-products/esuprt_ser_stor_net/esuprt_powervault/powervault-ml6000_Setup%20Guide2_en-us.pdf

The library, backup server and netapp filer are all configured within the zone on the switches

Switch 1 (tape drives 1, 3 and 5, Netapp port 0c and backup server port 0 connected)

zone name Z_WOKINGBKP_L0_TAPEDRIVE vsan 10
  device-alias WOKINGBKP_L0
  device-alias TAPEDRIVE_1
  device-alias TAPEDRIVE_3
  device-alias TAPEDRIVE_5

zone name Z_NETAPP3220_0c_TAPEDRIVE vsan 10
  device-alias BLK-NETAPP3220_0c
  device-alias TAPEDRIVE_1
  device-alias TAPEDRIVE_3
  device-alias TAPEDRIVE_5

sh zoneset brief
zoneset name WOK_FLEXPOD_A vsan 10
  zone Z_WOKESXi01_A_NETAPP3250A
  zone Z_WOKESXi02_A_NETAPP3250A
  zone Z_WOKESXi03_A_NETAPP3250A
  zone Z_WOKESXi04_A_NETAPP3250A
  zone Z_WOKESXi05_A_NETAPP3250A
  zone Z_WOKESXi06_A_NETAPP3250A
  zone Z_WOKESXi07_A_NETAPP3250A
  zone Z_WOKESXi08_A_NETAPP3250A
  zone Z_WOKESXi09_A_NETAPP3250A
  zone Z_WOKINGBKP_L0_TAPEDRIVE
  zone Z_NETAPP3220_0c_TAPEDRIVE

Switch 2 (tape drives 2, 4 and 6, Netapp port 0d and backup server port 1 connected)

zone name Z_WOKINGBKP_R1_TAPEDRIVE vsan 20
  device-alias WOKINGBKP_R1
  device-alias TAPEDRIVE_2
  device-alias TAPEDRIVE_4
  device-alias TAPEDRIVE_6

zone name Z_NETAPP3220_0d_TAPEDRIVE vsan 20
  device-alias BLK-NETAPP3220_0d
  device-alias TAPEDRIVE_2
  device-alias TAPEDRIVE_4
  device-alias TAPEDRIVE_6

zoneset name WOK_FLEXPOD_B vsan 20
  zone Z_WOKESXi01_B_NETAPP3250A
  zone Z_WOKESXi02_B_NETAPP3250A
  zone Z_WOKESXi03_B_NETAPP3250A
  zone Z_WOKESXi04_B_NETAPP3250A
  zone Z_WOKESXi05_B_NETAPP3250A
  zone Z_WOKESXi06_B_NETAPP3250A
  zone Z_WOKESXi07_B_NETAPP3250A
  zone Z_WOKESXi08_B_NETAPP3250A
  zone Z_WOKESXi09_B_NETAPP3250A
  zone Z_WOKINGBKP_R1_TAPEDRIVE
  zone Z_NETAPP3220_0d_TAPEDRIVE

Are we missing something from our zones?  How do I zone the robotic arm?  As far as I can tell on the ML6030 the only fibre connections were on the tape drives. We don't have the Fibre Channel I/O blades installed in the library.

c) We have the requisite LEO licenses.  Current qty is 5 which I believe is enough for the 6 tape drives in the library.

d) I don't think this is an issue.  This filer is not configured in a cluster and is just used as the SnapVault target for another set of NetApp filers.

By the way the requirement for the SSO was suggested by a Symanetc tech on phone support.  I just wanted to clarify my understanding of what was required which is why I raised a post on the forums.

I did manage to get backups to work when I added the NetApp filer as a file server via the Add wizard on the Backup and Restore tab.  However I believe this then meant that the backups were being done over the LAN and not via the Fibre.

Cheers

[Colin_Weaver]

a) Yes I did mean the backup exec server - we usually refer to the server that runs the core backup exec product as a media server

b) A library is made up of a machanism to move tapes from teh slots to the drievs as well as teh drives. This mechanism is often called the amr and bacuse of teh way BE send the job out to the "NDMP tape server" in each NDMP filer then each filer must be able to directly adderss the arm by havinn a connection. Where a SAN is involved the zones must be confiigured so that this is achievable as if teh arm cannot be address, the tapes cannot be moved from slot to drive and you will then end up with queued jobs waiting for tapes to be made availabel that will never happen.

On top of the arm connectivity, the arm itself feedsback how many drives are in the library and there are sometimes issues if BE cannot address all the drives from each "NDMP Tape server" Note I said sometimes, it seems to vary slightly from hardware implmentation to hardware implementaion as to whether partitioning the drives using zones causes issues or not. As such the exact answere to this bit can be "try it and see what happens"

With regards your posted zone list , sorry but I cannot comment if you have configrued that correctly as I am not familar with the exact ways to configure zones within a SAN,

c) Sounds like you have enough LEO licenses. Please be aware that licensing queries are handled by customer care and not tech support. That said I do not think that an invalid LEO count would cause your problem

d) I am not sure what we officially test with regards to using Snapvault from one filer to another and then trying to back that up

As the SSO requirements are sometimes confusing when combined with NDMP, we are already aware that some support technicians have in the past given the wrong advice and do take steps to avoid such problems. For your issue I would suggest you go back to the support case and ask (possibly by speaking to a duty manager) for an advancment to a more senior engineer to review the problem.

You possibly did backup over the LAN with your add server comment, however I would have expected you to need to disable the NDMP Protocol on the filer to do that. This is just because our remote agent uses a subset of NDMP and things conflict if you try CIFS backups with the hardware ndmp still enabled.

[Colin_Weaver]

Just for info I belive that you should expect to see the tape drives and library mroe than once in the BE Storage console when you try and configure this correctly and you do have to use the stgorage wizard to add both NetApp units into Backup Exec.

[G4v1n]

Hi Colin,

From discussions with the support guys they are telling me that it looks like the way that we currently have our tape library physically cabled will not work and that is what is causing our backup jobs of the NetApp filer to remain in the queued stage.

I am awaiting clarification but what has been suggested is that the tape library should be directly connected to the NetApp filer and then the BE server will send NDMP commands to the NetAPP which will perform the backup off to tape.

Essentially like the following which is detailed in the following symantec article:-

http://www.symantec.com/business/support/index?page=content&id=TECH48957

NDMP - Direct Attached:

In the direct-attached or local model, the backup server uses NDMP over the LAN to instruct the filer to start the backup. The filer then sends the data contained in the backup to a tape device attached directly to the filer via Fibre Channel or SCSI. During the backup, information about files and directories written to tape is transferred via NDMP over the LAN to the backup server where it is maintained in an index. For a restore operation, the backup server uses NDMP over the LAN to instruct the filer to start restoring files. Data is read from the filer's locally attached tape device and stored by the filer. During backup and restore, the backup server does not ever receive the actual data being written or read from tape. As a result, this model is commonly referred to as "LAN-free" backup.

Figure 1

Would you concur with what they are saying?

Cheers

[Colin_Weaver]

That is correct - but if you wire it only to the filer you can only write data from the NDMP filers to it and if you need to backup windows based data as well then the media server will also need to have direct access to the library - If you do not need to backup windows data as well, then you do not need the direct connection to the media server.

Of course if you also have more than one NDMP device/filer to backup than you can do a 3-way between the devices and only connect the library to one of the filers, OR you can wire the libary to multiple filers to send the data direct (which should be quicker) but does require that all of the filers can control the movement of tapes in the library (from slot to tape drive and back)

In your original description I thought you needed to share the library with multiple filers (to write filer owned data directly) and to the media server do do traditional remote agent backups of windows servers that do not store data on the filers. This is where everything in the library might need zoning to every device so that jobs can run on every host server (whether or not that host server is the Backup Exec media server or one of the filers)

EDIT: BTW in that tech article you quoted it does not mention the ability to share with an NDMP filer and the media server or even between more than one filer because in the version of Backup Exec that article was wriiten for that style of sharing was not supported. That came in with later versions and means that article does not describe all the scenarios that are now possible.

[G4v1n]

Hi,

Originally we would have liked to be able to perform backups of the NetApp filer via NDMP and also traditional remote agents backups from windows servers etc.

However this would have only been a short term requirement as we are in the process of moving all our exisiting servers into the Flexpod infrastructure.  Once that is done then the SnapVault configuration within our NetApps would take care of backups.  The requirement then would be just to offload the contents of the SnapVault Target filer onto tape for offsite storage for DR purposes.

The way we currently have it configured with the zoning the backup server and the NetApp filer can see all the tape drives but we get the job queued issue when attempting to do a NDMP backup of the filer.

So if I understand correctly rather than having the tape library, backup server and NetApp connected and zoned via the Nexus 5548 switch, it would be better to have both the backup server and the Netapp directly connected?  

To achieve this we would need to procure the Fibre Channel I/O blades for the Ml6030 tape library.

Alternatively it has been suggested that NetBackup maybe a better solution for our needs and we currently trying to get a trial license via our reseller.

I also noticed the article that I was directed to was for an earlier version so I checked in the Backup Exec 2012 Administrators Guide and found the following in Appendix N

About the NDMP Option
The Symantec Backup Exec NDMP Option uses the Network Data Management
Protocol (NDMP) to back up and restore Network Attached Storage (NAS) devices.
You can back up data from a NAS device to the following locations:
■ A storage device that is directly connected to the NDMP-enabled NAS device
(direct-attached)
■ A storage device that is connected to another NDMP-enabled NAS device
(3-way)
■ A backup-to-disk device on a Backup Exec server (remote)
■ A tape device that is attached to a Backup Exec server (remote)
■ A shared tape device or backup-to-disk folder that is connected to a SAN with
Backup Exec servers.

The feedback I have been getting from Support is that our configuration

1x Backup server with 2x fibre connections (1 to each Nexus 5548)

1x ML6030 with 6 tape drives (3x to each Nexus 5548)

1x Netapp FAS3220 filer with 2 fibre connections (1 to each Nexus 5548)

Nexus 5548 zoned as per the cut and paste in the previous comment 

does not look like it will work and the Filer should be directly connected instead.

I wonder if rather than 2 zones on each Nexus 5548 we should just have 1?  using switch 1 as an example

rather than:-

zone name Z_WOKINGBKP_L0_TAPEDRIVE vsan 10
  device-alias WOKINGBKP_L0
  device-alias TAPEDRIVE_1
  device-alias TAPEDRIVE_3
  device-alias TAPEDRIVE_5

zone name Z_NETAPP3220_0c_TAPEDRIVE vsan 10
  device-alias BLK-NETAPP3220_0c
  device-alias TAPEDRIVE_1
  device-alias TAPEDRIVE_3
  device-alias TAPEDRIVE_5

we would be better with :-

zone name Z_WOKINGBKP_L0_TAPEDRIVE vsan 10
  device-alias WOKINGBKP_L0
  device-alias TAPEDRIVE_1
  device-alias TAPEDRIVE_3
  device-alias TAPEDRIVE_5
  device-alias BLK-NETAPP3220_0c

Thanks