Is it the below configuration possible?
Master server has network 10.44 we (call it production network)
The Media server has 10.44 and 10.59 (the last one we call it a backup network)
The NetBackup clients only using 10.59, so it only has a communication to the media server
The production network and backup network are not routable to each other
Does the client need a direct communication to the Master server?
The client must have a connection also to your Master Server so you need to have also a 10.44 for the communication between client and master server.
Imagine if you dont have a connection between master and client, how will you configure a policy without knowing the client?
Thank you for your anwser and actually you already answerd mys supposed to be next version. Because currently we have some clients that user anothere backup network but it can only communicate to media server and not to master server and it is running fine, with your explanation now I clear why it can. Thank you once again. \\
@MarianneCan you give short explanation on how the scheduling works in this case ?
The scheduler is running the master which contacts the client with the backup selection as far as I know/remember from the process flow diagram
With normal filesystem backup (not ALL_LOCAL_DRIVES), the master server sends the backup instruction to the media server (with all relevant policy info).
bpbrm on the media server then connects to bpcd on the client.
bpcd on the client starts bpbkar that reads data and send it to bptm on the media server along with metadata to bpbrm on the media server. bpbrm on media server communicates back to master regarding job progress and catalog updates.
So, quite possible to perform backups with no access to master. (This is why users tell us regarding failed DB backups that 'there is no comms issue since FS backups are working fine.)
If this is Windows backup that needs VSS snapshot for WOFB, then the client also needs to communicate with the master as shapshot info is kept on the master.
With ALL_LOCAL_DRIVES in Backup Selection, the Master needs to connect to the client to run bpmount to determine file list.
The master then sends the backup instruction to the media server. The rest of the process is the same.
With database backup, the script initiates a connection to bprd on the master. The master server performs all the validation checks and hands over to the media server.
Updates to client log file is done by the master.
Same with user backup initiated on the client.
Client initiated restores (file or db) sends the request to bprd on the master. After validation, the master server hands over to the media server.
Can see I have been "hindered" by my background in database backups
The NetBackup Discovery Framework service (nbdisco) on the client sends discovered information to the master server, which stores it in the client database. There is discovered data for Oracle and SQL Server. I can't speak to that.
For Exchange, the discovered data is needed to resolve an IP-less DAG to a representative mailbox server. Discovered mailbox lists make GRT backups for Exchange 2013 and newer more efficient; otherwise the snapshot job has to gather the mailbox lists at backup time. For Exchange protection in a VMware backup, the discovered mailbox lists are needed for GRT rather than merely helpful.
Exchange GRT backup and restore needs connection from the client to the master server. The live browse and GRT restore processes (nblbc and nbgre) make bplist and bpclimagelist requests to the master server.
Restore from snapshot requires a connection from the master server to the client, because a client process (bppfi) acts as the media manager. Bpbrm on the media server starts bppfi, but bprd on the master server sends bppfi the restore file list.
Exchange and SharePoint database restore from a VMware image require connection from the master server to the client, because they also use bppfi.
Snapshot cleanup requires a connection from the master server (bppficorr) to the client (bpfis).
During snapshot creation bpfis on the client sends a "state file" to the master server for safekeeping. Other snapshot processes retrieve the state file information from the master server. Otherwise you have no DR capability for client snapshots.