Keep files in sync between Live Cluster and D/R Cluster

[Hoff]

If you want to roll your own cluster (which is where this whole discussion is headed) with this cluster on a budget approach, then you certainly can head that direction.

Sure, clustering and HBVS is extremely expensive.

At the other extreme of cost, couriers and off-site (encrypted?) backups can provide a (longer-latency) recovery process, too.

In the middle ground, that's rsync and nfs and distributed version control and such. And rolling your own.

Rolling your own clustering and your own DT means implementing and maintaining your own and testing your own scheme, and that gets expensive. (I remember when we all commonly wrote our own network stacks. We seldom do that anymore.) And implementing your own will be involved, as the error sequences are inherently difficult to test. Your site will be the first site through the error and failover sequences involved here, too.

I've implemented solutions here using both ways -- roll your own, and clustering -- based on cluster requests. The former is either a point solution for a specific failover case, or it's inherently far more work. As a solution provider, the latter provides better features and better economies for the customer.

But then, how much is an outage going to cost you?

Regardless of the solution you pick here, I'd recommend testing.

[Scott Langworthy]

To all:

Thanks for all the info. You've given me much to think about.

I will take this information and go do more research.

For now, I will close this thread, as we can go on all day about the different scenarios.

I appreciate all the feedback on the subject.

As always, Kudos to your expertise and advice.

Scott.

[Scott Langworthy]

See previous Email for comments.