Implement a method to identify the dead non-leader nodes/dispatchers.

[GoogleCodeExporter]

We have currently working version assuming that the non-leader 
nodes/dispatcher do not die. If they die, we don't have a mechanism of 
identifying and updating the node list/dispatcher list. So we need to 
implement a mechanism to do this.

Original issue reported on code.google.com by harshana05 on 2 Jun 2010 at 6:18

• Blocking: Updating the cluster ip list/dispatcher list when cluster nodes/dispatchers are dead. #10

[GoogleCodeExporter]

Currently I am thinking of sending a ping like a message to the current node 
list and checking the reply. If some node cannot reply, it will be removed from 
node list of all nodes. When this situation is identified by the leader, it 
will issue a restart or shutdown message by the leader.

Original comment by harshana05 on 12 Jun 2010 at 4:51

[GoogleCodeExporter]

I have implemented a basic version of this mentioned method and have tested. It 
is working. If we remove an active node by a mistake, it will automatically 
will be discovered by our Dynamic Discovery component. So no harm will be done 
in that case. I will test more and commit it to the codebase. So we can start 
looking at issue 10 as well.

Original comment by harshana05 on 12 Jun 2010 at 10:47

[GoogleCodeExporter]

Node Update service implementation implementation is complete. Tested and 
verified with node clusters and Dispatchers. 

Dispatcher Update Service implementation is commencing. Will be completed soon.

Original comment by harshana05 on 13 Jun 2010 at 10:52

[GoogleCodeExporter]

Implementation completed. Tested. Need to be tested a bit more. Marking as 
Fixed. 

Original comment by harshana05 on 13 Jun 2010 at 7:28

• Changed state: Fixed