cancel
Showing results for 
Search instead for 
Did you mean: 

DR Setup In an Non-Clustered Netbackup env with Full Catalog Replication

Abhijit_Srivast
Level 4

I have an environment where I have to replicate the catalog residing over SAN to the secondary site . 

Considerations : 

1) Have the same name and IP of the netbackup master 

2) DR and Primary will be build with the same patch level ( OS & NBU )

3) NIC's for DR will remain disable and will be out of network , build as an stand alone machine .

4) DNS Flush would be done 

5) 6- Databases to be replicated to the DR LUN ( */netbackup/db , */db/data , */volmgr/data , */kms , vault and var resp ) Page 58 HA guide .

Questions : ( I am referring to HA guide for Netbackup 7.5 )

1) Pg 57 pt 1 : Its asking to use a DNS alias name for master ( optimus01 is the master and domain is prime so optimus01.prime.com would help ? how ? ) since the plan is to use the same name while configuring .

2) pg 58 pt 5 ( unix ) soft link the following paths to locations on a common volume . Here the common volume is SAN LUN . So i am planning to create 6 volumes in the DISK GROUP of 3 TB LUN to match all data above in them than just linking . I am little messed up here :( 

3) pg 60 its asking to modify the vxdbms and server configuration files . What needs the modification and why ?

Please respond as this is a little urgent . 

Regards ,

Abhijit Srivastava .

1 ACCEPTED SOLUTION

Accepted Solutions

AAlmroth
Level 6
Partner Accredited

Hi,

Considerations:

1 & 3) You could certainly use a cold standby master as such with no clustering, and having the standby DR master offline. I guess you would need to connect to it out-of-band/other IP addresses.

2) is a requirement

4) I don't see the need for DNS flush as you use same IP. However, there could be need for some optional cache flush in NBU as well as network switch flush on ARP/redirection tables. We seem to always struggle with the builtin host cache in NBU though.

5) You should also make sure you have the most up-to-date bp.conf and other relevant touch files/config files not in db directory on the offline master as well, at all times. When in cluster, and making changed in the GUI, the GUI asks you whether you would like to make the change to all nodes in the cluster. Makes for less probability of mismatch on a failover.

Questions:

1) In your case you are not running a cluster, hence no need for a virtual hostname/IP address (it would however perhaps be more practical using it though)

2) soft linking is only relevant in clustering, as the disks are shared and exported/imported on the cluster nodes. I guess in your case, your SAN based mirroring/replication will provide the relevant LUNs in some state. You would need to make sure that the replica is consistent before mounting (pretty much automagically handled in VCS with supported replication agents such as VVR and SDRF etc).

3) Should possibly not be required as you use same hostname/IP address.

 

Just some initial points that pops out of my mind...

 

/A

 

 

View solution in original post

2 REPLIES 2

AAlmroth
Level 6
Partner Accredited

Hi,

Considerations:

1 & 3) You could certainly use a cold standby master as such with no clustering, and having the standby DR master offline. I guess you would need to connect to it out-of-band/other IP addresses.

2) is a requirement

4) I don't see the need for DNS flush as you use same IP. However, there could be need for some optional cache flush in NBU as well as network switch flush on ARP/redirection tables. We seem to always struggle with the builtin host cache in NBU though.

5) You should also make sure you have the most up-to-date bp.conf and other relevant touch files/config files not in db directory on the offline master as well, at all times. When in cluster, and making changed in the GUI, the GUI asks you whether you would like to make the change to all nodes in the cluster. Makes for less probability of mismatch on a failover.

Questions:

1) In your case you are not running a cluster, hence no need for a virtual hostname/IP address (it would however perhaps be more practical using it though)

2) soft linking is only relevant in clustering, as the disks are shared and exported/imported on the cluster nodes. I guess in your case, your SAN based mirroring/replication will provide the relevant LUNs in some state. You would need to make sure that the replica is consistent before mounting (pretty much automagically handled in VCS with supported replication agents such as VVR and SDRF etc).

3) Should possibly not be required as you use same hostname/IP address.

 

Just some initial points that pops out of my mind...

 

/A

 

 

jim_dalton
Level 6

AA's point about the data consistency is the your prime consideration out of all of the above.

How do you know - what is there in the design - that ensures, _guarantees_ data consistency, every time you exercise the DR environment?

That would worry me!

Jim