Re: Parallel Server for Failover...

The DLM and the PCM on each system will recover transactions and release locks from a failed node. Remember, OPS requires that all structures be available to both nodes, including rollback segments, redo-logs, etc. Each Instance in the OPS system is aware of the other instances operations and can handle 'clean' failures and nasty ones.

Given your description though, you may want to take a look at Oracles Hot Standby option. Essentially you run another instance on your secondary node in a perpetual recovery state. This instance does nothing but eat up redo logs from your primary system until you turn in 'up' as a full production copy of your primary system.

This type of recovery system works great, but you do loose any in-progress transactions, and the contents of the last un-archived redo log on the failed system. The benefit is that you don't have to deal with OPS, its difficult setup/admin and any performance impact.

Hope this helps,

-Kevin

David Sivick <dsivick_at_mindspring.com> wrote in article <5u48e5$6kb_at_camel12.mindspring.com>...
> Is anybody using parallel server for failover?
>
> What we have is an application that needs to be up 24/7. The front-end
> server is Forte and will have two servers up at any time. This
 eliminates
> the application server from going down. They can also switch databases
 if
> they don't get a connection to the master.
>
> We want to run parallel server, but virtually use it like a single
 instance
> server, with the second server only being used if the first goes down.
 This
> is going to be used on the floor of a manufacturing plant and the system
> being down can virtually stop the plant. (Don't even ask how I am going
 to
> handle maintenance/upgrades... can you say holidays! )
>
> I know I can use parallel server across the two machine CPU's, my major
> concern is, if I have a server that locks a block and/or exclusive lock
 on a
> row, and then the server crashes, how does the second (backup) server
 handle
> a request for that block? Since the first instance crashed, does the
 second
> still see a lock out there?
>
> Thanks,
> David
>
>
>
>
>
>
Received on Wed Sep 03 1997 - 00:00:00 CDT