Forum Discussion

jorixine's avatar
jorixine
Level 2
13 years ago
Solved

NetBackup 7.0.1 and Oracle with FailOver

Hi all,

 

I am trying to get configured a couple of RMAN scripts and NetBackup to run backups on my Databases.

The main problem I find is that NetBackup seems to rely on hostnames (or whatever you put in CLIENT_NAME in client's bp.conf).

 

The thing is that I have 2 servers, each one of them with one database, In case of a problem any of those DBs will failover to the other node. So the VIP or Virtual IP differs from the actual IP / hostname convination.

 

How should I configure Client / Server / Backup to be able to run the backups as expected? Remember I need to run other backup jobs on those nodes (for example OS file backups) and I cant configure the CLIENT_NAME as the VIP name.

 

I hope I made myself clear with the problem.

 

Thanks a lot in advance!

  • Keep local hostname in CLIENT_NAME entry in main bp.conf (/usr/openv/netbackup/bp.conf).

    Create bp.conf file in the Oracle user's home directory

    e.g:

     

    SERVER = <master>
    CLIENT_NAME = <virtual-name>

    See http://www.symantec.com/docs/HOWTO69052

     
    Another option is to hard-code virtual name in the script with NB_ORA_CLIENT parameter in the script, e.g.
     
    ALLOCATE CHANNEL ch00 TYPE 'SBT_TAPE' PARMS 'ENV=(NB_ORA_CLIENT=<virtual-name>)';
    ALLOCATE CHANNEL ch01 TYPE 'SBT_TAPE' PARMS 'ENV=(NB_ORA_CLIENT=<virtual-name>)';

    or

    ALLOCATE CHANNEL ch00 TYPE 'SBT_TAPE';
    SEND 'NB_ORA_CLIENT=<virtual-name>';

     

10 Replies

  • Keep local hostname in CLIENT_NAME entry in main bp.conf (/usr/openv/netbackup/bp.conf).

    Create bp.conf file in the Oracle user's home directory

    e.g:

     

    SERVER = <master>
    CLIENT_NAME = <virtual-name>

    See http://www.symantec.com/docs/HOWTO69052

     
    Another option is to hard-code virtual name in the script with NB_ORA_CLIENT parameter in the script, e.g.
     
    ALLOCATE CHANNEL ch00 TYPE 'SBT_TAPE' PARMS 'ENV=(NB_ORA_CLIENT=<virtual-name>)';
    ALLOCATE CHANNEL ch01 TYPE 'SBT_TAPE' PARMS 'ENV=(NB_ORA_CLIENT=<virtual-name>)';

    or

    ALLOCATE CHANNEL ch00 TYPE 'SBT_TAPE';
    SEND 'NB_ORA_CLIENT=<virtual-name>';

     

  • So the Backup Policy client should point to the 'name' that I have in /usr/openv/netbackup/bp.conf's CLIENT_NAME right?

  • Our DBA set up the RMAN script something like this

     

    if ${hostname} == "server1" then

      export $ORACLE_SID=sid1

      export NB_ORA_CLIENT=vip

    elseif ${hostname} == "server2" then

      export $ORACLE_SID=sid2

      export NB_ORA_CLIENT=vip

    fi

     

  • Oracle policy Client name should point to the VIP

    OS policy Client should point to physical hostname

  • Oracle policy must contain virtual name.

    Please see my previous post:

     

    Keep local hostname in CLIENT_NAME entry in main bp.conf (/usr/openv/netbackup/bp.conf).

    Create bp.conf file in the Oracle user's home directory with virtual name in CLIENT_NAME.

    Else, specify NB_ORA_CLIENT in script.

     

  • If I put the VIP name as de 'client' in the policy I get an error message stating that the server could not communicate with it. I can ping from both sides to each one of the names for each interface in both servers. It is all correctly configured via /etc/hosts.

     

    NB_ORA_CLIENT is specified as follows:

     

    send 'NB_ORA_POLICY=POLICY_NAME, NB_ORA_CLIENT=VIP_NAME';
     
  • Please show us exact error message and tell us where you see it.

    Ensure that entry vitual hostname exists in /etc/hosts of master server as well as media server(s).

    After hosts entries have been updated, clear NBU host cache on master and media server(s):
    bpclntcmd -clear_host_cache

    Please also ensure that these log folders exist:

    On master: bprd  (NBU must be restarted to enable this log)

    On media server: bpbrm and bptm (no restart needed)

    On Oracle nodes: bpcd and dbclient
    (dbclient must have 777 permissions)

  • The error after putting the VIP name as the Client in the Oracle Backup Policy is:
    access to the client was not allowed  (59)

    I can ping it from the backup server:


    [root@otasrv5 ~]# ping otadb1_uy_vip
    PING otadb1_uy_vip (10.9.3.155) 56(84) bytes of data.
    64 bytes from otadb1_uy_vip (10.9.3.155): icmp_seq=1 ttl=64 time=0.057 ms
     
    And I can resolve the server name and media center name from the client host (Master and Media servers are the same machine accessed by the same name, as far as I can see):
    [root@otasrv1 scripts]# ping otasrv5 
    PING otasrv5 (10.9.3.144) 56(84) bytes of data.
    64 bytes from otasrv5 (10.9.3.144): icmp_seq=1 ttl=64 time=0.058 ms
    [root@otasrv1 scripts]# ping otasrv5-mgt
    PING otasrv5-mgt (10.9.3.214) 56(84) bytes of data.
    64 bytes from otasrv5-mgt (10.9.3.214): icmp_seq=1 ttl=64 time=1.93 ms
     
    I just made a fast find search and I dont have any folder named 'bprd' I do have a binary, and seems to be running:
    [root@otasrv5 openv]# find . -name bprd -type d
    [root@otasrv5 openv]# find . -name bprd 
    ./netbackup/bin/bprd
    [root@otasrv5 openv]# pwd
    /usr/openv
     
    [root@otasrv5 openv]# ps -fe | grep bprd
    root      5035     1  0 Sep13 ?        00:00:08 /usr/openv/netbackup/bin/bprd
     
    I dont have any 'dbclient' folder or binary on the Oracle side.
     
    Backups on this particular node are working just fine via "hostname" (the name it is in CLIENT_NAME)
     
    Will make any difference if I put another bp.conf in the Oracle user home? I dont see how honestly, since the script autogenerated by NetBackup runs as root and then impersonates it self as oracle user by running su.
     
     

     

  • Log folders do not exist by default. They must be created under /usr/openv/netbackup/logs.

    Status 59 means that the active node is missing a valid SERVER entry in bp.conf.

    Check /etc/hosts entries on both cluster nodes as well as SERVER entries on both cluster nodes.

    Create bpcd log folder on both cluster nodes to troubleshoot status 59.
    Only bpcd log on active node should be written to when backup is attempted.