Hello Team,master server OS: Solaris 10Master server NBU version : 8.1.1we are planning to build our DR site. 1. we will build master server at DR site with same OS and version2. will install nbu binaries on newly build master server with same version3. we will configure all policies at DR site as inactive state4. we will create same db mount points (nbdb, nbemm) at DR site and that mount point will be in real time synch5. if our primary site goes down then we will re-name the secondary master server (same as primary master server) and run the show.6. if we need to perform any Test DR restore, in that case also we will re-name the secondary master server (same as primary master server) and perform or we can perform import of that particular image and do test restore.Is this plan looks good? kindly suggest. any other DR plan and suggestions are welcomed

I am afraid you won't be able to build the system by just QnA on the forum as there are a few important fundamentals you need to be aware of, they are quite extensively covered in the guide I've referenced earlier. Some of the key issues I see: The only supported way for the NBU Master to recover from a failure on a different machine is either running a continuous replication, sync or async that respects the write order fidelity, such as hardware-based replication or something like VVR. Rsync is only good if you shutdown NBU before starting the sync session otherwise you'd be replicating a bunch of non-recoverable garbage. If you read the NBU HA guide, it's got the setup procedure for the DR site and you have to perform it beforehand, not when the disaster strikes. You won't be able to install NBU and point it to a catalog replica, if you installing a new master in a DR scenario, you need a catalog backup. On the DNS name, I've already tried to explain that the best way to achieve it is through a manual or automatic (in cluster) assignment of an NBU virtual IP which the Master will belong to, whereas both nodes in PROD and DR will have their independent physical IPs, so it will be safe to move the virtual IP around without worries of IP conflicts. If you've configured the SLPs correctly and you have data on and connectivity to the respective media servers I see no issues with the recovery of data. In essence - I suggest to go through the HA guide in details. There are important fundamentals there.

The IP does not really matter but if you can create a dedicated virtual IP which you can bring up in the DR site or Prod site whenever required will help with clients whose DNS cache holds the name for a long time. You'd need to change the host IP/hostname for the prod Solaris box to "virtualise" NBU's hostname but it's not a big deal tbh. Have a look at Veritas NetBackup in Highly Available Environments Administrator's Guide and there "Deploying NetBackup master servers with full catalog replication" and specifically "About non-clustered NetBackup master server with catalog replication" - there is a good guide there

Kunalgoel - it really comes down to what you are trying to achieve in the DR scenario. AIR is a great option if you need to restore at the DR site and never run PROD backups from the DR site's master. Because AIR will not replicate backup policies and for all intents and purposes it's a different NBU domain, it just appears to be able to automatically import the replicated images and restore them if required in DR for DR clients (if you want to restore in PROD, for PROD clients, you really need to think how to deliver the data back to PROD) - so it's essentially a data DR tool, not an NBU master infrastructure DR tool. The only proper DR tool for the NBU Master infrastructure, if it's your goal - is the catalog replication, using clustered or non-clustered NBU master, depending on automation requirements.

Kunalgoel wrote: Hello, it really comes down to what you are trying to achieve in the DR scenario: I want that if our product site is down, then we can run the backups of all the clients who are configured on the product site from DR site, until our product site is live again.So that our backup is not lost till that time period. Secondly, When the prod site goes down, we can access all clients( All the clients are configured on the prod site) backup images from the DR site and if any restore request comes, we can do that from the DR site Means looking for HA of NBU Master server Ok this suggests me you need two things: HA for the master server, automated (clustered) or non-automated (non-clustered) with catalog replication. This will address the portability of Master's role between PROD and DR and the ability to perform backups and recoveries when the DR procedure is invoked. AIR will not help in achieving these goals as it does not cover the master role portability function, it's really data-only. Backup data replication between PROD and DR, this can be achieved by a proper configuration of SLPs to copy data between PROD and DR MSDPs. all these procedures are well covered in the guides referenced in this thread.

For the first question, yes you take XYZ and make it floating, basically that XYZ IP/Hostname (in reality you can have even a different IP, just keep the host name the same) will now belong to NBU and will alive on the active NBU master. You'd need to give another IP and hostname to the Solaris physical box and that can easily be XYZ-1/XYZ-2. To make the IP floating you add one more IP to the network interface through either interface alias or a virtual interface. When you need to move the master, you shutdown it in Prod and pop it up in DR.This is what a clustering software would do for you automatically but the same trick can be done manually and it's not rocket science. Now for the last question, while you can migrate from a standalone NBU install to clustered one, it's not free and there are no end-user tools available for it. This procedure requires a qualified and authorised consultant to perform a procedure through CatMan (proprietary tool) to modify the database and migrate to the cluster.

DR Site | VOX

23 Replies

Kunalgoel
Level 4
5 years ago
Hello,
Should we give same ip or different ip to DR master server?
By above method, if our Prod site goes down then we can perform client restore and run our infra from DR site?
- Mouse
  Moderator
  5 years ago
  The IP does not really matter but if you can create a dedicated virtual IP which you can bring up in the DR site or Prod site whenever required will help with clients whose DNS cache holds the name for a long time.
  
  You'd need to change the host IP/hostname for the prod Solaris box to "virtualise" NBU's hostname but it's not a big deal tbh.
  
  Have a look at Veritas NetBackup in Highly Available Environments Administrator's Guide and there "Deploying NetBackup master servers with full catalog replication" and specifically "About non-clustered NetBackup master server with catalog replication" - there is a good guide there
  - Marianne
    Level 6
    5 years ago
    Kunalgoel
    
    I agree with Mouse - the IP address does not matter, but the hostname does.
    
    When you install NBU, the hostname MUST be the same as the source master. It cannot be changed afterwards.
    But then, if you are planning on importing images, then the name does not really matter.
    If you want to do a catalog recovery (much better than import), then the hostname must match.
    Bear in mind that import of backups takes as long as the backup did.
    What is your RPO and RTO? Can you afford to wait for imports to happen?
    
    So, a DR plan needs to be structured around RPO and RTO.
    This will affect your budget for DR.
    As per the guide that Mouse referred to, catalog replication is good where the catalog at the DR site needs to be in sync with the Prod site.
    
    What type of backup storage are you using?
    If tape, will you be sending tapes that were written on a daily basis to DR?