Configuration of DD VTL with DDS for NDMP Backup

Question

We currently have a Problem configuring our DataDomain 860 as a VTL in Networker for the Backup of our VNX and VG2 NAS Storage.

Creating the VTL with NDMP-Devices in Networker was no Problem and as long as the Devices of the VTL are not shared we can successfully Backup a DataMover via NDMP to the VTL.

As soon as we add the other NAS-Storage Devices as shared devices to the VTL the Backup fails shorty after starting with multiple errors:

nsrd NSR info media warning: rd=ndmpserver:c176t0l8 (NDMP) reading: Input/output error

nsrmmd NSR warning NDMP Service Error: Tape close error with no tape opened

nsrd NSR info media warning: rd=ndmpserver:c176t0l9 (NDMP) moving: ndmp rewind failed, illegal arguments in request

nsrmmd NSR warning failed to close ndmp tape device, SCSI device is not open

nsrmmd NSR warning failed to open ndmp tape device, the device is in use

nsrmmd NSR warning ndmp tape setstate failed

nsrmmd NSR warning ndmp tape weof failed, illegal arguments in request

nsrd NSR info media warning: rd=ndmpserver:c176t0l6 (NDMP) moving: ndmp weof failed, illegal arguments in request

nsrmmd NSR warning ndmp header: I/O error

The only configuration changes we made from the default config is setting the Device Block Size to 512kB and adjusting the Target and Max Sessions.

Thanks alot in advance for any helpful input ;-)

AlexDerBasser · Accepted Answer

Hi everyone,

just to let you know the issue is now solved.

The error was not on the Networker-side and also not on the DD-side. VTL Configuration was also fine. The problem was that our VNX had a maximum of 4 concurrent data streams configured. Since we had configured a higher Client Parallelism the errors occured in NetWorker when the VBB Snapshots for the NDMP Backup were lining up. We now have the maximum supported number of concurrent Savestreams (8 for VNX) configured and also have set the Client Parallelism in NetWorker to 8 and configured 8 dedicated drives in the VTL.

Furthermore we had "Handler Default" and "512kB" configured for the NDMP Tape Device Block Size but only a maximum of 128kB is recommended by EMC which resulted in Read/Write Errors on Networker side (but no Backup failures) and multiple Buffer Overflow Errors on VNX side.

After making these changes the Backup now is running fine. Thanks to everyone for the help nonetheless.

ble1 · Answer

Did you test the paths already with server_mt from filer?

AlexDerBasser · Answer

I have loaded tapes to the drives and then tried

server_mt server2 -f tapeX status; where X = the number of the tested drive

That command was successful for both filers and all drives.

Was the zoning and Access Group Configuration we did correct?

We zoned both AUX Ports of both Filers to both HBA-Ports of our DataDomain and then created one Access Group on the DD with both AUX Ports of both Filers and the VTL (1 Changer; 12 Drives)

ble1 · Answer

I think the error you have suggests DDS error as it says device is already in use.  Since you use VTL, I wonder why do you use DDS?  You can (which is best practice with VTL) just create additional drives and assign them directly to filer.

AlexDerBasser · Answer

we currently only have one unlimited autoloader license free to use but some more dynamic drive sharing licenses. so we wenn for the configuration with one vtl and shared drives for both filers.

if i would like to give each filer dedicated drives i would have to create a vtl for each filer (and therefore need 2 autoloader licenses in total), correct?

ble1 · Answer

AlexDerBasser wrote:

we currently only have one unlimited autoloader license free to use but some more dynamic drive sharing licenses. so we wenn for the configuration with one vtl and shared drives for both filers.

if i would like to give each filer dedicated drives i would have to create a vtl for each filer (and therefore need 2 autoloader licenses in total), correct?

You can still use one library - you just add more drives to it which you dedicate to filer. No need to create separate library.

AlexDerBasser · Answer

Okay I now created a VTL with 14 Drives; 7 for each Filer and configured them without Drive Sharing in Networker

I have 2 Access Groups:

First AG: Filer1(all AUX-Ports of all DMs); Changer; 7 Drives

Second AG: Filer2 (all AUX-Ports of all DMs); 7 Drives

Unfortunately the behaviour is still the same; After starting the Group with both Filers he mounts the Tapes starts writing and after a short time the Error messages appear again.

ble1 · Answer

The order in which drives are listed should be the same as seen via sjirdtag - verify that first.  Then, via sjimm and server_mt verify tape device paths.  As drives are unique now, there shouldn't be ne reservation issues.

AlexDerBasser · Answer

Hmm okay. The Control Port of my VTL says: scsidev@176.0.0 (NDMP) But if i try to issue the command sjirdtag scsidev@176.0.0 on the Backupserver i get: scsidev@176.0.0: No such device?

ble1 · Answer

ok, I thought robotic was on backup server - it seems like it is on VNX (I usually never do that).  I'm not sure why device busy is given when VTL is enabled, but you may wish to check if this combo is in compatibility guide (I'm quite sure it is, but just to be safe).

ble1 · Answer

176 sounds awfully high.  Do you see is as 176 if you run inquire?  Note that some OSes (eg. Linux) have limitation in compiled version of driver (st in case of Linux) on how many buses they scan so some changed would be needed there.  Still, 176 sounds awfully high anyway.

AlexDerBasser · Answer

if I run

inquire -N Filer1 -T

on the Backupserver i get a list that doesn't show the VTL but says

Device 'c176t0l0' is busy. It cannot be opened at this time.

The Drives are shown on scsidev@176.0.1, 176.0.2...

If I disable the VTL and run the command again the Jukebox is shown:

scsidev@176.0.0:STK L180 0306|Autochanger (Jukebox), c176t0l0

S/N: 7260000000

ATNN=STK L180 7260000000

The corresponding Filer also Lists the VTL at 176.0.0:

server_devconfig server_2 -probe -scsi -nondisks

chain= 174, scsi-174 : no devices on chain
chain= 175, scsi-175 : no devices on chain
chain= 176, scsi-176
stor_id= celerra_id=
tid/lun= 0/0 type= jbox info= STK L180 0306
tid/lun= 0/1 type= tape info= IBM ULTRIUM-TD3 8711
tid/lun= 0/2 type= tape info= IBM ULTRIUM-TD3 8711
tid/lun= 0/3 type= tape info= IBM ULTRIUM-TD3 8711
tid/lun= 0/4 type= tape info= IBM ULTRIUM-TD3 8711
tid/lun= 0/5 type= tape info= IBM ULTRIUM-TD3 8711
tid/lun= 0/6 type= tape info= IBM ULTRIUM-TD3 8711
tid/lun= 0/7 type= tape info= IBM ULTRIUM-TD3 8711

chain= 177, scsi-177 : no devices on chain
chain= 178, scsi-178 : no devices on chain

AlexDerBasser · Answer

As far as I can see it is supported.

So to get the robotics to the Backupserver and test again we would have to create a new zone on the FC-Switches:

DD-HBAs to Backupserver-HBAs?

Then create a Access Group with the VTL-Changer and the Backupserver-HBAs

and adjust the other two Access Groups so they only Host the corresponding Filer and the Drives.

Finally reconfigure Library in NetWorker.

Am I correct so far?

ble1 · Answer

Yes.... to be honest, I always use robotics on backup server - even if I do not use any drives there.  Yes, your steps look right.

AlexDerBasser · Answer

Okay I did the zoning and Configuration on the DataDomain. Now the Robotics is handled on the Backupserver and each Filer has 7 Drives.

Unfortunately we still get the same error messages after a shot time.

Do I have to set the "in-use endpoints", "primary" and "secondary" endpoint on the data domain side to some specific value? use all or one specific endpoint? can all drives use the same endpoint?

Do I need to assign specific LUN-Numbers on the Data Domain side to the Changer/Devices? Currently the Changer has LUN0 and the Devices LUN1 - LUN14

NetWorker

Configuration of DD VTL with DDS for NDMP Backup

Was this post helpful?