create a simple package

[Liam Curtis]

Hi all,

Trying to create a very simple package that basically activates a volume group with relocatable ip. I think I am bit confused by failover parameters. We have oracle running under control of serviceguard with scripts that mimic those of our DBA, but we would prefer to have a simple script that simply activates volume group on primary node with no monitoring or automatic failovers as we have 24/7 dba's monitoring system (I would probably throw in simple shutdown halt cmd) Ive read thru admin guide and studied previous implementors script.

My plan is to:

1) use template with no services.
2) comment out SUBNET in to avoid failover on network outtage

We do not want ANY auto failovers to other node. Sounds crazy, but this is Oracle APPS we are talking about here. We would basically start cluster and once vg is activated, manually call the DBA scripts. Vice versa for halting cluster (though like I said, I should prob throw oracle shutdown script in the halt command)

Anything else I need to configure?

[Frank de Vries · Accepted solution]

Hi,
We also have oracle db's in package,
and although failover is possible,
it is not automated

In the package script we have
FAILBACK_POLICY MANUAL
NODE_FAIL_FAST_ENABLED NO
AUTO_RUN NO
LOCAL_LAN_FAILOVER_ALLOWED YES


RUN_SCRIPT /etc/cmcluster/picnew/control.sh
RUN_SCRIPT_TIMEOUT NO_TIMEOUT
HALT_SCRIPT /etc/cmcluster/picnew/control.sh
HALT_SCRIPT_TIMEOUT NO_TIMEOUT

I think the subnet does not necessarily
has to be commented out. As it just
specify what net to monitor, in case of
failire you can parameter it to take no
action.

We have no service defined.

Hope this helps

[Stephen Doud]

Liam, you are correct... do NOT list a SUBNET in the package configuration file if you don't want SG to cause a package halt if the NICs (primary/standby) on that subnet cannot transmit for whatever reason.

[siva baskaran]

there is no big proble check it netmask ip only. that might be u r ip series only

thanks
siva

[Liam Curtis]

Thanks all. It sounds like it should then work. Is there anything else in serviceguard like the above that would cause my package to failover automatically (like some setting somewhere or built in feature of MCSG)?

[Liam Curtis]

Hi Siva,

Thanks for replying. Not sure I follow your answer...

[Liam Curtis]

thanks for your help all. I also opened a case with HP, and it seems like if I set things as above, the only thing that would cause a failover is if the heartbeat stopped working between my nodes.