Re: agent, clusters and failover

Howard J. Rogers wrote:
> Are you talking about the Intelligent Agent?
>
> Or something else??
>
> If it's the Intelligent Agent, then of course you need one on every node,
> otherwise you can't manage the cluster *as* a cluster, but merely as a
> collection of separate instances. Enterprise Manager, for example, uses GSD
> on one node to talk to the other GSDs on the other nodes to perform, for
> example, cluster-wide shutdowns or backups or exports. GSD pulls off that
> trick by in turn talking to the Intelligent Agents on each node to handle
> the actual scheduling of a job. No Intelligent Agent, no cluster-wide
> scheduling.
>
> Maybe I've misunderstood your question though.
>
> For example, the question "should I bind the agent at the service or at the
> node" means nothing if we're talking about the Intelligent Agent, since that
> is intrinsically a "node process", running independently of whether or not
> the instances comprising your RAC are actually up or not.
>
> Regards
> HJR
>
>

Wild stab out of curiosity: Could it be that we're talking MS Cluster here? In that case, put the Intelligent Agent into the cluster group with the database, else you're gonna have a hard time with OEM. (Been there, done that...)

Regards,

Holger Received on Mon Mar 29 2004 - 02:05:59 CST