Re: 10g with HACMP (no RAC)?
From: Palooka <nobody_at_nowhere.com>
Date: Mon, 08 Sep 2008 19:22:21 +0100
Message-ID: <a3exk.184192$6s4.20312@newsfe14.ams2>
>
> No, afaik Oracle Clusterware and HACMP do the same things, i. e. defining
> and monitoring of cluster ressources. IIRC you can build a failover
> cluster with Clusterware as well as with HACMP. Only IBM support
> will be lacking. Since IBM and SAP are aggresively marketing DB/2,
> support for Oracle on AIX gets worse every day.
>
>
> Yes, it is. If HACMP decides that the active node has gone missing,
> it varies on the shared vgs, mounts the shared filesystems and sets
> an IP alias with the cluster's ip on the "public" interface of the
> standby node. It then executes all the application startup scripts
> which you have configured into the Ressource Group (RG). One of these
> scripts should start your Oracle instance, which will do an instance
> recovery. Or at least this is the case with a node failure.
>
> In case of a RG move your scripts should shutdown the instance on
> the activie side and restart it on the other node.
>
> Of course you have to do all the error checking in the scripts yourself
> and take care of every possible situation that could go wrong with
> db shutdown or startup.
>
Very many thanks, Uwe. To others, yes this may not be the optimal architecture, but as hpuxrac noted, the architectural decision has been made. I just wanted some guidance as to what will be involved on the Oracle side. It looks reasonably straightforward, provided, as Uwe says, the scripts are tested thoroughly.
Date: Mon, 08 Sep 2008 19:22:21 +0100
Message-ID: <a3exk.184192$6s4.20312@newsfe14.ams2>
Uwe Weber wrote:
>> Do we need to install any extra Oracle software (e.g clusterware)?
>
> No, afaik Oracle Clusterware and HACMP do the same things, i. e. defining
> and monitoring of cluster ressources. IIRC you can build a failover
> cluster with Clusterware as well as with HACMP. Only IBM support
> will be lacking. Since IBM and SAP are aggresively marketing DB/2,
> support for Oracle on AIX gets worse every day.
>
>> You mention the redo logs: I assume you refer to the online redo here, >> since starting the database on the second node should presumably be, in >> effect, the same thing as recovering from instance failure. That's >> right, isn't it?
>
> Yes, it is. If HACMP decides that the active node has gone missing,
> it varies on the shared vgs, mounts the shared filesystems and sets
> an IP alias with the cluster's ip on the "public" interface of the
> standby node. It then executes all the application startup scripts
> which you have configured into the Ressource Group (RG). One of these
> scripts should start your Oracle instance, which will do an instance
> recovery. Or at least this is the case with a node failure.
>
> In case of a RG move your scripts should shutdown the instance on
> the activie side and restart it on the other node.
>
> Of course you have to do all the error checking in the scripts yourself
> and take care of every possible situation that could go wrong with
> db shutdown or startup.
>
Very many thanks, Uwe. To others, yes this may not be the optimal architecture, but as hpuxrac noted, the architectural decision has been made. I just wanted some guidance as to what will be involved on the Oracle side. It looks reasonably straightforward, provided, as Uwe says, the scripts are tested thoroughly.
Regards,
Palooka
Received on Mon Sep 08 2008 - 13:22:21 CDT