Multiple Nodenames on the Same Client

[dhudson]

Hello,

I am setting up my nodes to backup to the TSM server and am using nodename_daily and nodename_weekly to seperate the nodes for different data retention periods.

I was wondering how one sets up the node for this. Do I have a seperate stanza in dsm.sys with a different nodename defined?

Actually, how do people normally do this? I'm starting to think it will be a lot easier to just define a seperate mgmt class. In which case, how would I schedule a server initiated backup to use the alternate mgmt class? There's no mgmt class option in the def sched command.

Thanks for any help.

 

[KB]

Take a look at backup sets to handle the weekly requirements. There is no need to do extra backups and juggle management clases.

 

[cyberdemon]

We use two different dsm.sys, each one with a different nodename and client port. The same as if you are using TDP.

 

[BBB]

Hi dhudson

Using two separate nodes is a pretty common way to go about things. The main alternative is backupsets or exports which can be good depending on how long they take to run in your environment. You can also use archives but this can increase your DB size pretty quickly and I wouldn't recommend it unless you have a pretty low file count & size.

You can't do this with a single node and by juggling an "alternate" management class.

To use multiple nodes for the one server
1. Add new node to dsm.sys (unix)
2. create a second dsm.opt file with new nodename
3. start new scheduler/dsmcad with -optfile=.... flag
4. create a new policy domain with a mgtclass with the retention period you want, and a dirmc management class as well (make it the same name as your usual dirmc mc name)
5. register the new weekly node in the new policy domain (and do all the usual password stuff)
6. create schedule and association for the new node

 

[dhudson]

Thanks for the help, I have a few questions to follow...

When you say add new node to dsm.sys, will it end up looking something like this...
------------------------------------------------------
nodename:root > cat dsm.sys
SErvername servername
COMMmethod TCPip
TCPPort 1500
TCPServeraddress servername
TCPNodelay yes
TCPWindowsize 64
TCPBuffsize 32
TXNBytelimit 2048
COMPRESSION NO
SCHEDLOGNAME /usr/local/log/adsmsched.log
SCHEDLOGRETENTION 7
ERRORLOGNAME /usr/local/log/adsmerror.log
ERRORLOGRETENTION 7
passwordaccess generate
nodename nodename_daily
inclexcl /usr/local/lib/inclexcl
schedmode prompted
LANFREETCPport 1500
LANFREECOMMmethod TCPIP
ENABLELANFREE yes

SErvername servername
COMMmethod TCPip
TCPPort 1500
TCPServeraddress servername
TCPNodelay yes
TCPWindowsize 64
TCPBuffsize 32
TXNBytelimit 2048
COMPRESSION NO
SCHEDLOGNAME /usr/local/log/adsmsched.log
SCHEDLOGRETENTION 7
ERRORLOGNAME /usr/local/log/adsmerror.log
ERRORLOGRETENTION 7
passwordaccess generate
nodename nodename_weekly
inclexcl /usr/local/lib/inclexcl
schedmode prompted
LANFREETCPport 1500
LANFREECOMMmethod TCPIP
ENABLELANFREE yes
------------------------------------------------------

And the dsm.opt file, what do you mean by "create a second dsm.opt file with new nodename"? There's no nodename specified in it. And do I need it if I'm doing the same backup?

Do I have to have the daily and weekly nodes in seperate domains? Is that to avoid having to specify a mgmt class in the options files? I ask because I will end up with 2 domains for UNIX and 3 for INTEL.

Thanks again.

 

[BBB]

something like that but use different log files. You need to distinguish the different entries in dsm.sys with a unique "servername <servername>" entry. The servername can be anything, its really just a name for that particular set of configuration, the actual server it uses is chosen by the tcpport and tcpserveraddress you set. Which leads me to:

You are right, when i said a second dsm.opt file with new nodename, i meant with the new "servername", as per the dsm.sys file servername entry you want to use.

 

[sgabriel62]

This is how I set mine up - to keep your logging separate - insert the correct path name for the dsmsched and dsmerror logs. Feel free to use extensions if you like

Steven

 

[dhudson]

And you say I use seperate domains for daily, weekly, monthly? I should seperate at that level?

 

[denisl]

You can keep it all in the same domain and bind your data via an include to the different management class (include /.../* WEEKLY_CLASS)

 

[BBB]

You can keep it all in the same domain and bind your data via an include to the different management class (include /.../* WEEKLY_CLASS)

Yes you can do that. But you will also need to make sure your DIRMC is also being handled for the long term backups or it will otherwise be kept for whatever the daily ones are being used for.

 

[ttrinh7975]

dhudson, if it's on WinOS. Create different Baclient folder for each node and dsm.opt that the node_name, also install TSM Scheduler Services for each node on the same machine.

Then associate the scheduler anyway you want on the server.

 

[Stephan]

dhudson, if it's on WinOS. Create different Baclient folder for each node and dsm.opt that the node_name, also install TSM Scheduler Services for each node on the same machine.

Then associate the scheduler anyway you want on the server.

Hi.

can you simple accomplish this by starting the WIndows GUI, going to the Setup Wizard and creating a new scheduler, BA client and web client? You just point it to a new dsm-2.opt for example and it will add a 2nd instance of each? Each opt must point to a different port? 1581, 1582?

Right now, i am running a Windows 2003 server with 1 node. I'd like to create a second node and separate my OS backup from my DATA backup...

Thanks.

 

[arjess]

Hi..

Yes,you can use the Setup Wizard to accomplish it and create one dedicated
scheduller to manage the schedule for the new define node.

 

[dhudson]

Take a look at backup sets to handle the weekly requirements. There is no need to do extra backups and juggle management clases.

I've just reread this and got my head round what was being said here. This sounds like a more sensible way to go about things.

I've just read the help file on generating backupsets, looks simple enough. You alluded to some performance considerations...?

 

[JohanW]

help generate backupset.

They are generated server-side, but it does take ages. A backupset always takes at least one full volume, no matter how small the backupset or how large the volume, and you cannot use backupsets with TDP data. (You can generate them, but you can't restore 'em.)

 

[Joker]

Hello,

The contents of the files dsm.opt dsm_daily.opt be the same or within dsm_daily.opt have to put the line 'nodename node_daily'?

Thanks for any help.

 

[Bharanij]

someone explain the 3rd point (3. start new scheduler/dsmcad with -optfile=.... flag ) ? on unix

 

[marclant]

someone explain the 3rd point (3. start new scheduler/dsmcad with -optfile=.... flag ) ? on unix

You resuscitated a 10 year old post, haven't seen these folks on here in a while.

It's best to start a new thread specific to your question and situation.

 

[Bharanij]

even its was a old post but i almost stuck in 3rd point and i dont know were to start in new post

 

[marclant]

You could start by explaining what you did so far and what you are trying to achieve. You have to help us help you. You haven't given us much to work with.