11-04-2012 08:26 PM
HI Experts,
We have a SQL cluster and as i know that if failover happens to another DB server the backup will fail as the netbackup cannot find the path of the files to be backedup , Is this true ??
Ex:- If i have DB01 and DB02 , if DB01 is failed over to DB02 then backup of DB01 fails with the error 71 (None of the files in the file list exist) ??
Please advise on this and provide a solution for this if i can still take a backup if the DRIVES are FAILED OVER to another DB server.
Thanks,
Nayab
Solved! Go to Solution.
11-05-2012 04:59 AM
In simple terms, your physical boxes (nodes) will probably have one or more static addresses, they never change regardless of the state of your cluster and then your sql instances will each have floating addresses and these can relocate from one physical box to another as the state of the cluster changes.
Hence to backup an instance you use the instance address in the policy which will exist on a node...you dont know which node at the time of backup and you dont need to know: the floating address is part of the cluster framework and is managed by it.
So as per MvdB and others: to backup your OS: use the node addresses: the physical addresses of the machines in the (windows) policy; to backup the sql instances use the floating addresses in the (sql) policy (or virtual addresses as other have called them) and associated bch file.
jim
11-04-2012 08:35 PM
11-04-2012 08:47 PM
Node name in the sense how it should be given i mean let me tell how it is in my environment
It is a Five Node Cluster DB01 ...DB05 on WINDOWS 2008 and all the HOST name will be like INDB01,2,3,4,
If i configure virtual host name will the backup happen in the case of failover as well ?? If yes how to do it please advise
Thanks,
Nayab
11-04-2012 08:58 PM
Are you referring to NODE name or CLUSTER NAME ???
11-04-2012 09:07 PM
ALL SERVERS ARE PHYSICAL WINDOWS 2008 R2
Netbackup Version :- 7.1 On all Master , media , clients
11-04-2012 09:26 PM
We normally create 2 policies for clusters:
1. Policy for physical nodenames to backup local drives, e.g.
C:\
Shadow Copy Components
2. Policy for Virtual hostname to backup shared/cluster data.
This can be a application/database policy type to backup database online or MS-Windows policy to backup database dumps.
For a FileShare cluster, this policy will be the drive letters for shared data.
11-04-2012 10:07 PM
All our machines are PHYSICAL ones so the first point applies here now please let me know wat information i have to provide for u to check and make the backup happen when it was failed over to other SERVER ??
11-04-2012 11:59 PM
11-05-2012 01:13 AM
All our machines are PHYSICAL ....
Please speak to your Cluster Admins to explain the concept of Virtual hostname and Virtual IP.
This is how users and other applications are connecting to SQL.
The whole point of a cluster is to be able to connect to the database/application regardless of which physical nodename the cluster is active on.
This is where a Virtual IP and hostname is used. It fails over along with drive letters and SQL services.
Please get this virtual hostname and IP address from Cluster Admins.
Ensure forward and reverse lookup between master and media server(s) and the virtual hostname and IP.
Create a separate policy as per no. 2 in my post above.
11-05-2012 02:20 AM
Hello Maria,
I have checked with my CLUSTER ADMIN he gave me one IP ADDRESS and HOSTNAME and said this is our CLUSTER IP and HOSTNAME ( VIRTUAL ) , please advise me how to proceed on this do i need to create a new policy for this else i can add this IP to existing policies as we have 4 Policies for DB backups like Filesystem , FULL , LOG , etcc.... Please Advise.
Thanks,
Nayab
11-05-2012 02:41 AM
It all depends on what exactly is in policies.
If Filesystem is for local drives only (e.g. C:\ and SCC) then one policy for the physical nodenames.
All policies that are backing up SQL data must be changed to Virtual hostname.
If you are using the SQL agent, you need to update the script(s) as well.
See the section in NBU for SQL Admin Guide as per Yasuhisa's post.
11-05-2012 04:59 AM
In simple terms, your physical boxes (nodes) will probably have one or more static addresses, they never change regardless of the state of your cluster and then your sql instances will each have floating addresses and these can relocate from one physical box to another as the state of the cluster changes.
Hence to backup an instance you use the instance address in the policy which will exist on a node...you dont know which node at the time of backup and you dont need to know: the floating address is part of the cluster framework and is managed by it.
So as per MvdB and others: to backup your OS: use the node addresses: the physical addresses of the machines in the (windows) policy; to backup the sql instances use the floating addresses in the (sql) policy (or virtual addresses as other have called them) and associated bch file.
jim
11-14-2012 03:06 AM
Will i be able to Use the CROSS MOUNT BACKUP to have the drives backed up when it actually fails over to other DB server ...may i know hw it works and also if any impact.
11-14-2012 04:59 PM