MCservice Guard change of package script

[Rogelio Ordinario]

Do I still need to shutdown the cluster if I am modifying the package control script (customer defined run/halt commands). Why and Why not?

[Christopher McCray_1]

Everything that I can think of says no. The reason is that you are only making changes to the package and not the cluster. For example, the dba group here needed to change the name of their listner for oracle. I made the chnge in the oracle.sh file. All they had to do was rename it in their listener file and stop/start the listener. The next time the package stops and starts, the new information will be read in.

Hope this helps.

Chris

[Uday_S_Ankolekar]

Hello,
If you change the control script before halting cluster, next time during failover service guard take the new script.
Process I would opt for , Halt the cluster, change the controlscript and then restart cluster

Good luck
USA

[John Poff]

Hello,

I've made plenty of changes to my control scripts without bringing down the cluster. Uday has a good point, though. If you want the current running package to halt next time using the old script, you'll have to wait and bring down the package first. Otherwise, you should be fine.

JP

[Mladen Despic]

Hi Rogelio,

I don't think you need to halt the cluster in order to
reconfigure a package. However, you will probably want
to HALT THE PACKAGE before changing your
control script. This is recommended by HP since "timing
may cause problems."

Regards ... Mladen

[linuxfan]

Hi Rogelio,

You do not necessarily need to shutdown the package/cluster when you modify the run/halt commands. HP recommends it but i have had no problems modifying the run/halt commands.


http://docs.hp.com/cgi-bin/onlinedocs.py?mpn=B3936-90045&service=hpux&path=../B3936-90045/00/00/51&title=Managing%20MC/ServiceGuard

-HTH
Ramesh

[Sanjay_6]

Hi,

My opinion would be that you wont have to shutdown the cluster if you are making changes to the package control script. This is because the package control script is called from the package configuration file and hence does not affect the cluster binary. You will though have to replicate the change in package control script to all other nodes where this package will get activated in case of node/package failover.

You can even make changes to the package configuration script and replicate them when the cluster is running using
cmapplyconf -v -P /etc/cmcluster/pkg_name/pkg_acii_conf_file.

IT is recommended that if you are making any changes to the package ascii configuration script and intend to apply them, you should shuutdown the package before that. But the Cluster can still be operational. In both cases.

Hope this helps

Thanks

[Sridhar Bhaskarla]

HP recommends to halt the package for changing the run/halt scripts. However, I modified my run/halt scripts many-a-times without getting into a problem. Policies like failover, failback, autorun etc can be done online.

-Sri

[melvyn burnard]

There are a few things to bear in mind here.
Which version of SG? 10.06 and below I would not make any changes without halting the cluster first.
What changes are you making?
Keep in mind that changing the run/halt script should not have too much impact on th epackage, and definetly nothing on the cluster, you should naot have to halt the cluster if you are running 10.10 of SG or above, but I would recommend halting the package first, make the changes (not forgetting to copy them to the other node(s) ) and thene TEST! start and stop the package on each node.
Most issues occur where changes are made, and then not tested, or tested on only one node.
HTH

[Santosh Nair_1]

Although it is recommended that the package be halted when making changes to the package script, I think its safe to make the changes on the fly.
You would need to halt the package if you were making changes to the package configuration file, because this gets gened into the cluster binary file, but the control script are just scripts which get invoked as necessary and so can be changed on the fly.
Just make sure to synch the scripts across all your nodes in the cluster.

-Santosh