Tru64 UNIX oddities

[js88699]

I'm hoping someone with more Tru64 experience than I have can shed some light on a situation I just found myself in.  Before I ask my question, I'll describe the environment the best I can.

1)  I have two nodes (lets call them nodeA and nodeB) that are part of a cluster running Tru64 OSF1 V5.1 alpha.

2) The NetBackup client is 6.0mp4 (yes I know its end of life soon) and the master and meda servers are 7.1.0.2

3)  Both nodes see the exact same storage.  For example /usr/openv/netbackup/bp.conf is the exact same file seen on both nodes.  Not copies of each other, the EXACT same file.  Having an OpenVMS backup ground, this sort of cluster is not new to me, but it is on a UNIX platform using NetBackup.

Here is where things got interesting.  Over the weekend nodeA crashed with a hardware problem.  On Monday morning they started the database application up on nodeB.  They wanted to make sure the DB backup was going to work.  Not a big problem, or so I thought.  Long story short, after jumping through some hoops I got the DB and OS backups to work and things were good last night.  One of the things I found was the CLIENT_NAME was commented out in the bp.conf file.

Today the vendor came in and fixed nodeA.  The application folks moved everything back to nodeA.  They sent out an email to let me know.  So I said I would move the backups back.  I changed the client name in the DB poilicy and the DBA put the RMAN script back.  When I ran the backup, I got an error 239.  This was not at all what I was expecting.  So I logged into nodeA to make sure the DBA did not typo something in the script, it was OK, I checked bp.conf and found it had nodeB in it and a modification date from yesterday.  Since this node was down, that puzzeled me.  That was when I found out about how the storage is configured.  Both nodes see all the storage at the same time.

Right now I have CLIENT_NAME = nodeA in the bp.conf and tested sucessfully the OS and DB backups.  The OS backups use the cluster name (node) so that appears to work with no issues.  Since they both see the same storage only need to run the backup on one of them.  Even with the backup policy set to use the client name node and the bp.conf set to use nodeA it worked.  I was expecting that to fail with a 239 but it worked.  This got me to thinking..

What is the effect of having the CLIENT_NAME commented out of the bp.conf?  I don't know if this is something unique to Tru64 or not.

[revarooo · Accepted solution]

CLIENT_NAME
Specifies the name of the client as it is known to NetBackup. There can be one CLIENT_NAME entry and it must match the name used in the policy that is backing up the client. The only exception is for an alternate client restore, where the name must match that of the client whose files are being restored. (See “Allowing Redirected Restores” on NetBackup Configuration Options 486 NetBackup DataCenter System Administrator’s Guide - UNIX page 309.) The client installation procedures automatically set CLIENT_NAME to the value specified on the ftp_to_client or install_client command in the installation scripts. You can add this option to the /usr/openv/netbackup/bp.conf file on a UNIX client. It can also be added to a $HOME/bp.conf file on a UNIX client but this is normally done only for alternate-client restores. If the value is not in any bp.conf file, NetBackup uses the value returned by the gethostname() library function. 

[js88699]

Revaroo,

Your last sentence is what I was looking for.  As you say it SHOULD match the name in the policy, but I ran one expecting it to fail with a 239 and it worked.  bp.conf had "nodeA" in it, the the policy just had "node".  "nodeA" is not in the policy at all.  This has nothing to do with any restores.  I'm tempted to comment the CLIENT_NAME out again like it used to be, but as long as it's working I'm wanting to leave it be and change it if we have to run from the other.