How to confiugre tape library in Linux master server

[NikitaC]

Hi All,

Please help me to configure/install the new tape library on netbackup linux master server .

Netbackup version :- 7.5

Master server Os :-Red Hat Enterprise Linux Server release 6.7

Tape Library ;-  TS3310

Tape library is connected to switches and zoing has been done. also driver directory is not available under /usr/openv/volmgr/bin .

[root@######### sbin]# /sbin/lsmod |grep sg
ipmi_msghandler        38701  2 ipmi_devintf,ipmi_si
sg                     29318  3

[root@######### bin]# ./tpext -get_dev_mappings_ver
device mappings version in the EMM database is 1.115
device mappings version from the local file is 1.115
Local device mappings file is up-to-date

Tape library is not showing in  cat /proc/scsi/scsi output.

ThankYou .

[RiaanBadenhorst]

If the library is not visible under /proc/scsi/scsi then the OS cannot see it. If the OS cannot see, NetBackup cna't see it.

Check your zoning again.

[WVT]

Work from library backwards. Library and drives online and ready, nameserver of switch shows devices and hba of host online, zoning correct? Check HCL, I believe IBM atdd drivers are preferred.

[Marianne]

I agree with above posts. Forget about NBU and first get the OS to see devices. NBU needs the OS to see and use devices.

[mph999]

The order in which things should be seen (for libraries and drives)

(1)  HBAnywhere or Sansurfer HBA utility if installed.  Not that many people seem to have this installed,but it's well worth it.  The utility depends on the make of card, QLogic or Emulex (and I can never remember which utility goes with which brand).

(2)

OS ...

/proc/scsi/scsi for Linux

Device Manager for Windows

ioscan for HP-UX  etc

... and so on (differs depending on OS)

(3)

scan command (will show both libraries and drives)

Until (1) works, forget (2).  Until (2) works forget (3)

There is no point running tpautoconf if scan doesn't work, it's kinda the same command anyway.  There is even less point in running tpconfig as this only shows configured devices - not what is actually there (put another way, I can make devices appear in tpconfig that doen't actually exist if I wanted to ...)

Usually, if scan shows the devices, they will then configure correctly, however, scan must show correct output, for example, device files etc ...  As I always mention, although Veritas supplies the scan command it is NOT a NBU command as such, it will run and provide output even with NBU services stopped.  NBU does use the scan command (to configure devices) but the scan command does noy use NBU.  So if scan isn't working, it's a problem at the OS or scsi layer.

In your case (2) doesn't work (no devices seen in /proc/scsi/scsi - so this is either a OS / SCSI layer or zoning issue most likely.  It could be a library fault but probably unlikely.  It would be useful to have (1) available, the HBA utility as this is independant of the OS (well, kind of) and effectivly allows you to see if the HBA can see the devices, which is why it is so useful.

[WVT]

No, if switch doesnt see, nobody sees.Library web console first , check drive and robot status then ask for nameserver output of switch that shows all logged in devices. This library is also partitionable into multiple logical libraries. Robot and drives must show successfull login to fabric as N-ports at correct speed...best to not use auto and hard set..depending on device/switch firmware mix this will get you 10-15% higher tape drive throughput . Zones activated in running config, then OS/HBA.

[sdo]

Good advice from WVT - hard set port speeds - if one spends just two minutes thinking about it - one should realize why this is good advice. . Hey, I think the advice to check SAN nameserver might be lost on some, but is true purist - if you want to build/check upwards fromphysical to topolofgy to server, which obviously takes physical server HBA in to consideration, then this is the best advice I've seen for a long time re checking actual SAN connectivity - if one proves name server entry visibility, and zoning visibility (think fcping command) then any OS connectivity issues can only be due to the server driver layer upwards. Hence the further advice for a detailed check of compatibilty matrices and vendor inter-op. . I'm no expert, but I know enough to know that you know what you're talking about. Pleas stick with us, we could use some additional expert advice in some areas.

[winter_prince]

Hi Nikita,

mph999 is correct on this...

Forget netbackup for now.. Please set the basics right first....

• Is the SAN zoning done correctly ? Please clear that first.
• Is the Server allowed access to the tape library? You might have to do this from Tape management console.
• Once above steps are cleared, Please verify from the OS level:
  
  Check which FC Port are connected and logged in:
  
  for i in `ls -1 /sys/class/fc_host/ | sort`
  do
  echo "$i  `cat  /sys/class/fc_host/$i/port_name` `cat /sys/class/fc_host/$i/port_state`"
  done
  
  Force scan of the available ports: 
  
  # echo "- - -" > /sys/class/scsi_host/host3/scan
  # echo "- - -" > /sys/class/scsi_host/host4/scan
  
  
  Install rescan-scsi-bus.sh tool for scanning the interfaces.
  
  # rescan-scsi-bus.sh
  
  Check whether the tapes drives are available from the OS now.
• run netbackup scan command and see whether the devices are available.

[NikitaC]

Hi All,

SAN zoning is fine .

[NikitaC]

Nameserver output of switechs

[Will_Restore]

rescan the scsi bus or reboot the host.

[Will_Restore]

ok, two drives.  where's the changer(robot)? 

[WVT]

Use wwpn for drives, not wwnn. Why are the drives NL , should be point to point fabric N port. ...Set at library console.....Are these switches interconnected?

[WVT]

These are LUN 1 robots if I remember correctly. You wont see it in name server. It bums off of drive.

[WVT]

If these switches are NOT interconnected, his zones are wrong. Look at nameserver. Hba that ends in e2 at port id 010200 is in zone with tape drive on a different switch. 01 is switch domain...each switch is different...02 is physical port ....00 is alpa address. Tape drive has port id beginning with 02...different switch. If the switches are interconnected the zones are okay, except for the incorrect usage of wwnn for tape drives and NL showing.

[NikitaC]

The swithces are interconnected.

I have rebooted the server but still library is  not visible on OS .

So I am going to arranging the field engineer visit to check the library connectivity and to configure the remote console access .

Hopefully this will help. 

[WVT]

Was this library used somewhere else? What hba are you using? Driver loaded? Zones need to contain drive wwpn not wwnn. Drives should be set to login as Fabric port. Show output of dmesg command right after rebooting.

[MG1989]

Hi NikitaC,

Other than the points mentioned by other boarders just to add one more point. Check whether this library comes with Host access configuration.    If so you have to register the host (server) after zoning and provide access to the drives and robot.

Ignote if this library does not have that feature or it is disabled.

Thanks.

[Marianne]

Yes. Linux needs sg and st drivers.

See the Linux chapter in NetBackup Device Configuration Guide:  http://www.veritas.com/docs/000003098 

[NikitaC]

Hi All,

My apology for responding you late.....

Finally I got Library access and got the wwpn from library managment console, unfortunately my tach was giving me wwnn :(

Again we have performed zoning using wwpn number and now I can tape drives and medium changer on OS ....

Also yes switches are interconnected.

device name is not showing .....