Hi,

On Mon, Dec 29, 2008 at 05:38:02PM -0800, Patrick McShane wrote:
>
> Hi,
>
> I am a heartbeat newbie and have a question regarding Oracle-SE 10g 
> (standard edition) used in a two node active/standy heartbeat V2.x 
> configuration.
>
> Our Oracle license limits what features are available to use with the 
> oracle-se 10g.   We don't have OCFS or Dataguard.   Our option is limited 
> to using the Oracle configured as master on the primary heartbeat node 
> (using the OCF oracle RA) with an available hot standby Oracle database on 
> the secondary heartbeat node.  
> When Oracle is running in standby mode, redo/archive logs are rsynced (via 
> cron) from the master node every 15 minutes or so and applied to the Oracle 
> standby by DB on the secondary node.   Technically the Oracle standy DB as 
> a defined resource needs to up and running on the the secondary heartbeat 
> node so that archive logs can continue to be applied which keeps the 
> standby DB within 15 minutes of the master at any given time.
>
> Now for the real question.   Most services in heartbeat are defined as UP 
> or DOWN.

There are also three-state resources, also known as master-slave.
That has to be supported by the resource agent.

> How can we setup the failover senario for the Oracle DB such that when the 
> failover actually occurs, we can first change the mode of the already 
> running Oracle standby DB to master mode on the secondary heartbeat node 
> after applying the latest available archive/redo logs? 
>
> The Oracle master DB is typically running on the primary heartbeat node and 
> the Oracle standby DB is typically running in standby mode on the secondary 
> heartbeat node.
>
> As far as I can tell, the /usr/lib/ocf/resource.d/heartbeat/oracle script 
> is just designed to start or stop the DB without any hooks for doing 
> something in addition to STARTING or STOPPING the DB.

Right. There's just an option to manage the database backup mode.

> We need to be able to run some SQLPLUS scripts just prior to changing the 
> mode of Oracle DB to master and then starting the listener.
>
> Any recommendations/feedback on this would be welcome.      Maybe adding 
> some hooks to the oracle RA?  

It looks like the "standby" database is actually a running
instance. Right? In that case the oracle RA will report it as
running. No hooks can help here.

Probably the only way to handle this is a master-slave resource
agent which would be able to recognize the "standby" and "master"
states as well as to promote or demote an instance.

Thanks,

Dejan

> Regards,
> Pat McShane
>
>
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to