Community
Participate
Working Groups
Currently the fail-over management involves manual interaction with the FailoverParticipants, in particular the call to backup.setType(MASTER) after the former master has become unavailable. We need to provide a FailoverMonitor network facility that coordinates multiple FailoverParticipants (2 or more) by communicating to a FailoverAgent per FailoverParticipant. The FailoverMonitor is also supposed to automatically manage the connection details of the backup synchronizers and all the clients in the network, so that all nodes only need to know the FailoverMonitor connection details.
New monitoring framework: org.eclipse.emf.cdo.server.internal.net4j.syncing.FailoverMonitor org.eclipse.emf.cdo.server.internal.net4j.syncing.FailoverAgent Updated examples: org.eclipse.emf.cdo.examples.server.FailoverExample.Monitored.Monitor org.eclipse.emf.cdo.examples.server.FailoverExample.Monitored.Agent1 org.eclipse.emf.cdo.examples.server.FailoverExample.Monitored.Agent2 Committed to HEAD
I've moved FailoverMonitor and FailoverAgent to the public org.eclipse.emf.cdo.server.net4jpackage because the agent is abstract and needs to be subclassed.
Clients can now also participate in the monitored failover scenarios by using a new session configuration type: CDONet4jUtil.createFailoverSessionConfiguration(String, String, IManagedContainer) The sessions created by this configuration will detect server/network failure (as usual) and (new) query the failover monitor for the new master connection details, connect to that master and reregister all existing views with the new master. An example is in org.eclipse.emf.cdo.examples.server.FailoverExample.Monitored.Client.main(String[]) Committed to HEAD
Available in R20110608-1407