TSM Client on Linux-Cluster

[max]

hi,

i can't find docs on tsm client installation for a linux cluster (RedHat 4).
I have installed the client in a Microsoft Cluster and a HACMP, but this is the first time on a linux box.

Can you please give me some advice ?
Is the procedure similar to HACMP ?

The server is at 5.2.10.0, I'm trying with 5.3.0.0 client

looking forward to your answers
cheers
max

 

[moon-buddy]

Yep, similar to HACMP and Windows clustering.

You need to have 2 TSM schedule services running on each node - one for the local and one for the shared making sure that the shared resource scheduler handles the fail over properly.

 

[max]

... which makes 4 for each node, because i'm saving to a TSM Disaster Recovery Server as well ...

thank you
max

 

[max]

ouch ....

just tried a "dsmc q ses SE=CLustername"

and it gives me

ANS1481E The CLUSTERNODE option is not supported on this system.

is it compatible with Linux ?

max

 

[moon-buddy]

Nope. The CLUSTERNODE option is a Windows option.

For Linux, you need to define basically a multi-stanza dsm.sys. Portions of dsm.sys will point to the local node while the rest points to the shared node.

The service would then use portions of the dsm.sys to start up the scheduler for the local and shared resources.

Just not sure how the hand-over works when the nodes failover. Anyone?

 

[chad_small]

Usually there is a "heartbeat" process running that looks for the active side of the cluster and if it does not see it it starts up the TSM process on the alternate node. You code try and use AIX's scripts and tweak them for Linux but it wont be supported by IBM.

 

[moon-buddy]

Chad,

Thanks for the info. I wasn't clear if there is a way to fully automate fail overs in Linux. It is easy in Windows using the Cluster Manager.

Nice to know that there will be a starting point by using the AIX scripts.

 

[max]

ugh,

"it won't be supported by ibm" ?
How am i supposed to backup a linux cluster then ? I used the same procedure i once did on a hacmp, and the "clusternode" option is a must !

please give me a hint
max

 

[moon-buddy]

As I mentioned earlier, you need a multi-stanza dsm.sys file like:

SErvername <TSM_SERVER_NAME
COMMMethod TCPip
TCPPort 1500
TCPServeraddress 192.168.2.2
NOdename <LOCAL_NODE_NAME>
Exclude "/opt/tivoli/tsm/log/dsmsched.log"
COMMRESTARTDURATION 15
ERRORLOGRETENTION 5 D
MAXCMDRETRIES 6
RETRYPERIOD 10
SCHEDLOGNAME "/opt/tivoli/tsm/log/dsmsched.log"
ERRORLOGNAME "/opt/tivoli/tsm/log/dsmerror.log"
SCHEDLOGRETENTION 14 D
SCHEDMODE PROMPTED
TCPCLIENTADDRESS 192.168.2.10
PASSWORDACCESS GENERATE
INCLUDE <local_files_to_backup>

SErvername <CLUSTER_NODE_NAME>
COMMMethod TCPip
TCPPort 1500
TCPServeraddress 192.168.2.2
NODename ticoonprodb
TCPCLIENTADDRESS 192.168.2.10
TCPCLIENTPORT 1501
COMMRESTARTDURATION 15
ERRORLOGRETENTION 5 D
MAXCMDRETRIES 6
RETRYPERIOD 10
SCHEDLOGRETENTION 14 D
SCHEDMODE PROMPTED
PASSWORDACCESS GENERATE
SCHEDLOGName /opt/tivoli/tsm/log/dsmclustersched.log
ERRORLOGName /opt/tivoli/tsm/log/dsmclustererror.log
INCLUDE <shared_files_to_backup>

When you start the dsmc schedule service for the cluster, do:

/usr/bin/dsmc schedule -se=<cluster_node_name>

Also, don't forget the dsm.opt file:

SErvername <local_node_name>
Domain /
SErvername <cluster_node_name>
Domain /whatever

The only point I am not sure would be how to handle the fail over automatically. Chad had pointed out to the AIX scripts.

 

[max]

hi moon-buddy,
i know about the mulistanza dsm.sys file, i'm ok with that. You have to realize though that the clusternode option is essential, because tsm would "confuse" the versioning and retention stuff, should the cluster switch to another node. TSM must know that there is a cluster and use the same retention and versioning policy, as it were a single tsm node. That is exactly how it works on Windows and AIX. The problem I asked for first is that there is no documentation on redbooks guides for linux clusters, and i was wondering if the HACMP documented procedure would work, which now i'm afraid it doesn't.

Still i'm asking how am i supposed to backup a linux cluster, has anyone done it before ?

max

 

[max]

hi,
still wondering if someone has coped with this problem before. Did anyone succeed to backup a linux cluster ?
looking forward to your response
thanks
max

 

[wildwest]

There is no formal clustering support in the Linux client. If you can make it work sufficiently for you, that's great, but don't call support talking about a cluster setup because they aren't going to hear it. If you can explain your problem in terms that don't include clustering, that may work.

You might be able to backup each machine in the cluster separately. There should be some way for you to make a script that decides which machine is down, and then the "up" machine can start backing up whatever is shared under it's own node name, although the first backup will be huge. But something's gotta give. Since clustering isn't formally supported something will suffer somewhere.

I've never set it up though, so I'm not sure what exactly to do.

 

[ravikanth]

Hi All,

Thanks a lot for posting your valuable solutions. I am looking for the same thing.

Ravikanth

 

[max]

hi,

well Ravikanth, considering how long i've been waiting for ... good luck.
It seems nobody backups linux clusters ...

cheers
max