Proper way to achieve failover and recovery in symmetric clu

[scheerer]

I seem to be having problems figuring out how to recover a live node in a symmetric cluster after failover.

Scenario:
2 live nodes (A, B) with corresponding backups (C, D) running in self discovery broadcast cluster.

Client using standard JMS/JNDI interface for ConnectionFactory. (not the core api)

I've ensured that while running the example that I end up having consumers on both nodes A and B. I then kill node A (kill -9) and the failover to node C works as expected... no hiccups whatsoever.

The problem is... how do I bring recover node A without causing any service interruptions? (I'm running 2.0.0-BETA5)

[timfox]

Recovery is explained in the user manual.

[anthonylee]

Hi,

I guess you are asking : how to bring back the original live server(node A) without shuting down the running back server(node C)?

I am also thinking of it.

[scheerer]

I think I understand the recovery piece in the manual, but it requires shutting down the backup, which would cause a service interruption to those consumers/producers connected to the backup at that time.

Is this not correct?

Also, when trying to bring a backup/live pair back up into the cluster after failure i continually see these errors about duplicate ids.

[DiscoveryGroupImpl] There seem to be more than one broadcasters on the network broadcasting the same node id

Is this expected?

[timfox]

"scheerer" wrote:
I think I understand the recovery piece in the manual, but it requires shutting down the backup, which would cause a service interruption to those consumers/producers connected to the backup at that time.

Is this not correct?

That is correct, for 2.1 we will add functionality to sync a new backup with live without interruption.

[anthonylee]

HI Tim,
I feel interest in this 'live sync' feature.
When will the date to release 2.1?