pkg failover----Oracle installed in internal HDD

[Rgomes]

Hi All,

We configured 2 node cluster on 2x rx5670 server, it is a new setup. Scenario is:



1)Oracle 9i is installed in internal HDD on both rx5670. ( /dev/vg00/lvol7 ------/u01)

2)Databaseâ s mount points are in SAN va7110.



Configuration is ok upto now, package failed over to another node if we pull-off lan cables from any server. But it is not happening when Oracle user is connected with the pkg( unicorn_pkg). Control log shows package and respective node is halting but never halts down. It seems like it hangs.



Is it for, Oracle is installed in internal HDD? Cannot free-up or reset user connection to another nodeâ s Oracle?



Thanks in advance for kind suggestions,

Rich

[Massimo Bianchi]

Hi,
unfortunatly i see no logs.

Your problem is not with oracle on local/shared disks but with oracle itself.


But one common situation is that oracle is closed with a simple shutdown.

Better solution is closing oracle with a "shutdown immediate", this will prevent further logon to oracle, or better to use the oracle integration mcsg script.

I do not remember if it is on pay or free, but the logic is as follows, and you can always implement it with scripts :)

- issue a "shutdown immediate" and for a process with a timer.

- when the timer elapses, check: if oracle is up, then issue a "shutdown abort" and exit.

The startup script must used some recover, to allow for a correct startup. Usually this is implicit in oracle9.


Be sure to have latest strems patch, otherwise you will have some oracle process that won't die even with a "kill -9", and that can prevente the correct switching of the package.


Massimo

[Rgomes]

Hi Bianchi,

Thanks for your reply.
At the time of package failover, what will happen to users those who are connected with Oracle/package IP, will they be disconnected?
Or their connection remain the same with virtual IP( i mean, package IP)?

richard

[Rgomes]

HI,

This is the pkg log. I halt the primary node( dtcellm) while connected as 'oracle' user and I got disconnected after halting down the dtcellm node.

Pls see and give your feedback.

Thanks,

RIchard

[Massimo Bianchi · Accepted solution]

Hi,
the log of the shutdown is perfectly neat.

Users will be sconnected, because the virtual IP address is moved from one server to the other.

But when they reconnect to the virtual IP, they will see no difference.

MCSG gives you the ability to offer always the same virtual IP address, but it is up to the upper layer of tcp/ip to maintain the communication and control it. it is found on some application, like the SAP AS (application server), where you have a reconnect option. But if you are using pure sqlplus or similar, there is no chance.

Being disconnected during a package failover is absolutely normal, because the ip is switched.

Massimo

[Rgomes]

Hi Massimo,

Thanks for the explaination. It is clear now.

regards,
Richard

[Rgomes]

Forgot to give you point, :(

richard