cancel
Showing results for 
Search instead for 
Did you mean: 

robot control host in cluster

Jan_Voskar
Level 3
Hi,
I am experiencing a strange behavior of NetBackup , when configured in cluster ( VCS running on two solaris nodes ). After installing NBU6 on both nodes, running cluster_config script and configuring devices on both nodes, everything is OK except one thing.
The node, on which I first run inventory of robot is chosen as robot control host for all tapes in robot instead of virtual name, so after failover of NBU to other node, I cannot do inventory of robot ( the error is - cannot connect to vmd ). Everything is working except inventory.
When I delete all tapes and run inventory from command line ( from any node ) with parameter -rh virtualnameofnbuserver, media list is populated with correct (virtual) robot control host and it is working after failover to any node of cluster. Am I doing something wrong during configuration of NBU into cluster? Did anyone meet with this problem ?
Thank you.
12 REPLIES 12

Peter_Jakobs
Level 5
Partner Accredited Certified
Are you using the Java interface or the Windows Administration console?
I've seen it an a Tru64Unix cluster using Java. How is your bp.conf and vm.conf file?

Peter

Jan_Voskar
Level 3
Hi,
I am not using Windows console, I configured it from solaris gui (java console).
The hostnames of cluster nodes are ujo and teta. Virtual hostname of NBU master server is nbumasta.
bp.conf file is same on both nodes except CLIENT_NAME ( one is ujo and other teta :(
SERVER = nbumasta
CLIENT_NAME = ujo
EMMSERVER = nbumasta
VXDBMS_NB_DATA = /opt/VRTSnbu/db/data
SERVER = teta
SERVER = ujo
CLUSTER_NAME = nbumasta

vm.conf is same:
CLUSTER_NAME=nbumasta
DEVICE_HOST=nbumasta

Chia_Tan_Beng
Level 6
Hi Jan,

While login to java console, do you use the actual hostname or the virtual hostname? In your case, nbumasta should be used.

Jan_Voskar
Level 3
I thought about this earlier and even when I use nbumasta ( virtual name ), robot control host is still one of the local hostnames ( ujo or teta ) :(

Chia_Tan_Beng
Level 6
Although its the same physical host, the host name you entered while login into GUI does make a different. I experience that before but not sure whether you're having the same problem but worthwhile to try. You may also having problem in the Netbackup Management -> Catalog when attempting to verify or duplicate image from the GUI.

Yasuhisa_Ishika
Level 6
Partner Accredited Certified
I guess the robot is registered on global device database(globDB) as its robot control host is teta.
Can you delete this robot once, and configure again? Make sure the device host and robot control host are both nbumasta.

Jan_Voskar
Level 3
Hi,
1.
I tried verifying of images aftrer failover to the node which is not robot control host and it is OK ( so I suppose duplicating also ).
2.
In vmglob DB there is no robot control host assigned to robot. Here is a short output from vmglob -listall command and as you can see there is a dash in robot control host row:
device type: robot
device name: ROBOT0
hostname: nbumasta
volume database hostname: nbumasta
robot number: 0
robot type: TLD
robot control host: -

When I cofigure robot through wizzard, the only option is to choose local node ( ujo or teta ) as DEVICE HOST for robot, otherwise it would not be possible to add a robotic path definition. The same apply for manual robot ( and drives ) configuration from GUI or tpconfig.

3.
More and more I think it is not a configuration problem of my installation, but a bug of Java console, which assigns a bad name ( local not virtual ) during a robot inventory, because as I wrote in my first post, when I run robot inventory from command line and supply a virtual name as robot control host , it is OK.

Thank you for all posts.

Yasuhisa_Ishika
Level 6
Partner Accredited Certified
Well, globDB seems to be correct.
I hope I could help you, but I have no idea so far.

Bartosz_Kabala
Level 2
HI
I have this same trouble on aix and vcs and netbackup 6.0 did u find any solution?

Jan_Voskar
Level 3
Hi,
yes I have a solution (or workaround).
When you are doing first inventory of robot, you have to do it in command line with vmupdate command and you have to use a switch -rh with virtual name of netbackup server, not local hostname of server.
If you already have some tapes assigned, you have to move "ALL" the tapes to standalone (in GUI for instance) and then run vmupdate -rh "virtual_name" along with other options of course.
As soon as there are tapes assigned to virtual robot control host you can use GUI for inventory of robot.

Bartosz_Kabala
Level 2
thx for help - vmupdate works fine but...
now when i choose inventory i can see correct device host (name of cluster) but in robot i can see inactive node of cluster !!! and after switch cluster i can see inactive node of cluster which earlier was active. do u know why?

Jan_Voskar
Level 3
Hi,
I face the same problem ;-). It is probably a bug of java gui ( there are more ), but still you can do an inventory in GUI when you click Media in GUI under Media and Device Management and in the upper menu choose Actions->Inventory Robot