Forum Discussion

vyruzj's avatar
vyruzj
Level 2
9 years ago

Media Server switching during DR

Hi,

I am running NBU 7.6.1.2 on Windows 2012 and i have a Master server in the DR site and 2 media servers (one in production and one in DR). I am thinking a way to switch the media servers to use for backup during DR.

My environment:

During usual business days my production servers: SvrA, SvrB and SvrC will be backup by my media server MediaSvrA in the production site. The servers SvrA, SvrB and SvrC is using Microsoft Clusters and it will failover to DR when there is a disruption to the services in production environment.

Since the failover is automated the microsoft clusters, is there a configuration that can be done in an automated way so that the failed-over servers will use the media server - MediaSvrB in the DR for backups? If not then the servers will try and will backup to MediaSvrA.

I know that i can manually switch to use MediaSvrB in the policy configuration, but sometime server administrators will not inform backup administrators that the server has failed-over and causing the backup traffic to flow across the WAN.

7.6
Backup and Recovery
NetBackup
Windows Server (2003-2008)

10 Replies

Sort By
  • RiaanBadenhorst's avatar
    RiaanBadenhorst
    Level 6
    9 years ago

    Only way to do it is to write your own script to test which site the clients are active, and then update the policies accordingly using bppllist.

  • Michael_G_Ander's avatar
    Michael_G_Ander
    Level 6
    9 years ago

    You might be able to use a failover Storage Group for this, if the MediaSvrA is unavailable when the clusters failover to the DR site or the option failover media server.

    Another but very messy way, is to have an alias for the policy media server, that points to either MediaSvrA or MediaSvrB depending on the site. Not even sure this works in the newer versions of Netbackup, has not done anything like since 3.4/4.5

  • RiaanBadenhorst's avatar
    RiaanBadenhorst
    Level 6
    9 years ago

    You risk backing up across the WAN though if Netbackup determines that the "local" stu is unavailable.

  • areznik's avatar
    areznik
    Level 5
    9 years ago

    All the places I've worked at handled this kind of thing by throwing resources at the problem. We would back up all of the cluster nodes, no matter if they were active or not. PROD site cluster nodes would be configured to go to PROD site media servers and DR site cluster nodes to DR site media servers. ALL_LOCAL_DRIVES backups all around. If a failover happened PROD site backups got smaller and DR site backups got bigger. When it came time to restore it was pretty easy to tell which was the active node at the time. 

    Its kind of wasteful to back up passive nodes every day when there's almost no changes, but with deduplicated storage its really not too big a deal. 

    For this to work your cluster would need to be using VIPs/Cluster Names on top of server hostnames. So for example SrvrA.local.com is the hostname of the PROD cluster node you backup, and SrvrX.local.com is the DR cluster node you also backup. The cluster DNS name that all the applications and users use to access the app running on this cluster is App.local.com. When theres a DR scenario App.local.com name fails over from SrvrA.local.com to SrvrX.local.com along with other cluster resources like storage, dbs, webservers, etc. Since you're backing up both SrvrA and SrvrX you dont need to do anything to start picking up the failed over filesystems at their new home. 

     

    Some questions for you:

    Are the PROD and DR masters separate environments or is it a clustered master install? 

    Are the cluster nodes you're backing up VMs or physical?

    What kind of policies are backing them up VMWare (VIP) or MS-Windows? 

    Are they always running (passive) in the DR site or only spun up if DR is invoked?

  • vyruzj's avatar
    vyruzj
    Level 2
    9 years ago

    Sometimes the application cluster failover to DR due to some resource issues (network, disk, etc) but the media server is perfectly fine. So i think that in this case the failed-over application cluster will/might backup to the media server in the production site.. 

    I will read up on the app_cluster option. Thanks!

  • vyruzj's avatar
    vyruzj
    Level 2
    9 years ago

    I am not a very good scripter. But i'll try to get some help from team mates from another department. Thanks!

  • vyruzj's avatar
    vyruzj
    Level 2
    9 years ago

    Sounds like a possible solution. Might need to test it out as i am currently backing up using the cluster names. So when comes to restoration, the list available for restore might not be available for the cluster name as i am using the physical hostname for backup. And I am not using ALL_LOCAL_DIRVE directive, a backup policy is configure for each mount point (15 of them) as each of the mount point is 3TB (easy to re-run a job if anything fails).

    As for your questions:

    Are the PROD and DR masters separate environments or is it a clustered master install? 

    [Me] There is only 1 master server and its sitting in the DR site managing backup policies for both Prod and DR.

    Are the cluster nodes you're backing up VMs or physical?

    [Me] They are physical file server with up to 15 mount points with 3TB each.

    What kind of policies are backing them up VMWare (VIP) or MS-Windows? 

    [Me]MS-Windows

    Are they always running (passive) in the DR site or only spun up if DR is invoked?

    [Me] They will be running passive in DR. Being a file server for a manufacturing industry, availability is main priority hence whenever there is an issue to the file server in PROD, immediately failover to DR so that user can still access their document and files.

     

     

  • areznik's avatar
    areznik
    Level 5
    9 years ago

    What kind of storage are the 3TB mount points on? SAN, NAS, Local?