Continental Clustering and OPFS for DRP ??????????

[Henrique Silva_3]

We have an application and want to setup OPFS ( OPS and MCSG for HA - high availability ) for HPUX.

For the DRP solution, we looked at CA ( continuos access for the SAN XPs ), but the pricing for setting it up is too great.

So, I am thinking of setting up Oracle Standby DB or replication using Libelle or some third party tool.

Now, when installing MCSG with the OPS option, we noticed that you can not have continental clustering and OPFS (???????).

Does anyone have run into this problem ? We want to setup DRP between two sets of systems connected via OPFS. If that is not possible, then, we will need to setup replication between the HA systems.

Anyway, any help will be appreciated here.

Cheers,

Henrique

[melvyn burnard]

Unfortunately at this time, Continental Cluster does NOT support using ServiceGuard OPS Edition, so I am afraid you wil have to rethink your plans.
What is the distance between sites? do you need Continental Cluster?

[Henrique Silva_3]

We are planning to replicate the data from Atlanta to Tampa ( 1,000 miles ???) or Atlanta to Houston ( 2,000 miles ??? ). It all depends of the new network infrastructure we will be getting.

My understanding of metropolitan vs. continental clustering has to do with the ability to own the lines ???? as opposed to leasing them , or is this campus clustering ?

In any case, OPS would work fine for the HA rate would be higher than active/passive MCSG systems that would switch over in case of failure.

Now, does metropolitan clustering works with OPFS ??? What is the issue if you do the switch over manually ? Can't yo uestablish a reliable heart-bit between the boxes ? And as long as the two instances are the same, via replication of some sort, what is the issue or difference from continental clustering and MCSG on a non-OPFS instance ???

Cheers,

Henrique

[Henrique Silva_3]

I guess, because you can have 2 or N boxes, it would be very hard to determine if the INSTANCE is down !!! Maybe that is the reason.

What if we implement MCSG to use the Oracle client to establish the heart bit, as opposed to having a heart-bit between the boxes ? That way, we could tell whether the instance is up or not, independently of system statuses.

Cheers,

Henrique Silva

[melvyn burnard · Accepted solution]

OK that eliminates MetroCluster, as the current limitation is 100 kilometres separation.
OPS is NOT supported by Continental Cluster, so you would have to use normal ServiceGuard and Oracle, and then use the Oracle Standby replication facility to back up te database to the remote cluster.
Take a read of the manuals at:
http://docs.hp.com/hpux/ha/#ContinentalClusters

[Henrique Silva_3]

Thanks Melvin.

Back to the drawing boards
:-(

Cheers,

Henrique Silva