Forum Discussion

AlexeyF's avatar
AlexeyF
Level 4
2 years ago

NetBackup configuration with multiple isolated backup vlans

Hello dear community

We've inherited a NetBackup infrastructure that was designed and installed by previous company with lack of documentation.

I'm trying to figure out why MS SQL intelligent policy backups don't work and what should be the correct configuration of server and client names because it seems that the cause is there.

Master server has multiple network interfaces with multiple ip addresses and multiple names:

master.admin.domain - 10.1.1.10

master.prod.backup.domain - 10.20.1.10

master.preprod.backup.domain - 10.21.1.10

 

Client has 2 interfaces

client.admin.domain - 10.2.2.22

client.prod.backup.domain - 10.20.1.22

On the server in bp.conf

SERVER = master.admin.domain
SERVER = media.admin.domain

On the client the following servers are indicated in the Registry: 

master.admin.domain

master.prod.backup.domain

media.prod.backup.domain

 

As noticed, client uses his admin interface to communicate with master server admin interface during the initial stage of the backup but it fails then with not a huge amout of details in logs.

 

17 nov. 2022 00:00:25 - Error bpbrm (pid=232595) from client client.backup.domain: ERR - exit status: <2>
17 nov. 2022 00:00:25 - Error bpbrm (pid=232595) from client client.backup.domain: ERR - bphdb exit status = 2: none of the requested files were backed up

If we add a following line to the hosts file on the client the backup works:

10.20.1.10 master.admin.domain

but it does ot seem as the right solution. Moreover, on the second site, with (what it seems to me) the same configuration of master/client, backup works without errors.

Backup with script (mssql batch file also works on this client)

Could you please advice what should be the algorithm to debug the error and is the network configuration (vlans, hostnames) correspond to BP?

thanks in advance

Linux
  • Nicolai's avatar
    Nicolai
    Moderator
    2 years ago

    Hi AlexeyF 

    On the SQL client you should put master.prod.backup.domain as the first server entry, since they are both on the same subnetwork (10.20.1.10)

    If you on the client ping :

    master.admin.domain - 10.1.1.10

    master.preprod.backup.domain - 10.21.1.10

    Do you then get a response ? if not, those interfaces should not be on the server list on client.admin.domain. Only list interfaces who actual works

    Also something to be aware of working with multihomed systems:

    • Each server and client has a internal routing table. Start by understanding how it works (netstat -rn on Windows) and how it can be manipulated (route statements). 
    • Each system has a default gateway where all network request are send, if host isn't connected direct to the same VLAN
    • You need to know subnet and their subnet mask,  what looks like to be a "local" (non-routed) VLAN may be a network behind a router depending on the subnetwork mask.
    • Always Use FQDN names (e.g master.preprod.backup.domain not master).
    • Be aware of the domain suffix search list, it sometimes explain why "master" and be resolved but not e.g. master.prod.backup.domain 
    • AlexeyF's avatar
      AlexeyF
      Level 4
      2 years ago

      hi Nicolai 

      thank you for your response.

      I tried to put the name master.prod.backup.domain and media.prod.backup.domain. Actually it seemed logical to me, but it didn't work neither.

      I had a session with Veritas support and they told me that as on the master server the name configured in bp.conf is master.admin.domain, the same name should be used on the client. Don't know if this is true or not.

      So I decided to come back to the configuration that I find on all the clients, ie master.admin.domain in the first place.

      client can ping both interfaces on master server : admin and prod.backup. Prod backup is an isolated vlan so to communicate it uses dedicated route in its routing table ad to communicate with master.admin it uses default route which is via his admin interface. nothing exotic here.

      So if you advice to switch to master.prod.backup.domain I could try. Do I need to regenerate certificates for the new name?

       

      Regards,

      Alexey

       

      • Nicolai's avatar
        Nicolai
        Moderator
        2 years ago

        AlexeyF 

        Not sure I can add any good advice here, sound like everything in place.

        A good tool for tracing traffic is wireshark or tcpdump on Linux/UNIX, it may help you understand how Netbackup traffic flows. Just trace on 1556, 13724 or 13782

  • AlexeyF's avatar
    AlexeyF
    Level 4
    2 years ago

    Nicolai

    have you ever tried this configuration with MSSQL Intelligent Policy ?

    It seems that it doesn't want to use the alternative name master.prod.backup.domain

    In the generated batch file (by SIP) i see NBSERVER is still master.admin.domain

    Tried to follow the recommendation in this article NetBackup™ for Microsoft SQL Server Administrator's Guide (veritas.com) but SIP still uses master.admin.domain (the name found in server's bp.conf)

    So can't figure out how to make master.prod.backup.domain work as the master server for SQL agent.

    • davidmoline's avatar
      davidmoline
      Level 6
      2 years ago

      Hi AlexeyF 

      For the troubleshooting did anything change when you added the SQL Server agent registry value to use the preferred master (master.prod.backup.domain) and ALSO set the USE_REQUESTED_MASTER = FALSE (this second change is required to change the precedence).

      You say this is working on a different site - are you sure none of the above have been done there?

      I was going to suggest the addition of a local host entry to change what the client thinks is master.admin.domain to be the same IP as master.prod.backup.domain - but I see you don't like that as a solution.

      Cheers
      David

  • AlexeyF's avatar
    AlexeyF
    Level 4
    2 years ago

    davidmoline 

    >> did anything change when you added the SQL Server agent registry value to use the preferred master (master.prod.backup.domain) and ALSO set the USE_REQUESTED_MASTER = FALSE (this second change is required to change the precedence).

    I've tested once again. Nothing has changed :\ MSSQL server still trying to use master.admin.domain despite USE_REQUESTED_MASTER = FALSE

    In the generated by Intelligent Policy batch file I see the line NBSERVER master.admin.domain

    >> You say this is working on a different site - are you sure none of the above have been done there?

    On the different site the configuration is using master.admin.domain as a master server name and it works...