failover config for nodes with centeraSDK

How many access nodes do you have configured on the primary and replica centeras?

What application are you using?

I would recommend initially that you include in the connection string all the IP addresses of the Access Nodes in the primary, at least a minimum of 2 and specifying more than 4 probably isnt really necessary as if 4 Access Nodes are down the likely hood is that the entire cluster has a problem but include as many as you feel comfortable with.  Their order is not important.

You do not have to specify the replica IP addresses, the SDK will auto discover them and will failover any read attempts when the primary becomes unavailable.  Writes DO NOT failover and the application will receive an error if the primary has failed and a write is attempted.

The next question to consider is what do you want to happen when the application starts and the primary cluster in unavailable.  If you or your application vendor just want to stop any writes (which is fine the policy engine can try and do the archive the next day when hopefully the primary is back online) then you don't need to consider anything else.

If however you want the application to use the replica cluster for writes if the application discovers the primary is unavailable on start up, then include the replica IP addresses in the connection string AFTER the primary IP addresses.  On start up the SDK will timeout on the primary IP addresses and use the replica IP addresses.

As long as you have bi-directional replication set up, when the primary comes back on line replication will sync both clusters.   After the primary is back online you will need to restart your application so that the application starts writing to the primary again.  Replication will ensure that any recent clips written while the primary is down can still be found.

Note that there is no automatic in flight write failover to the replica in the SDK.

You probably should discuss this with your application provider to ensure they don't have a problem with this.

Also there is a whitepaper on Centera replication on powerlink at

http://powerlink.emc.com/km/live1/en_US/Offering_Technical/White_Paper/H5553-emc-centera-replication-wp.pdf?mtcs=ZXZlbnRUeXBlPUttQ2xpY2tTZWFyY2hSZXN1bHRzRXZlbnQsZG9jdW1lbnRJZD0wOTAxNDA2NjgwMzVjMThhLGRhdGFTb3VyY2U9RENUTV9lbl9VU18w

it hasnt been updated to the 4.1 rev of C* but the information in there is still relevant.