[CONTROLLER-782] Clustering : Applying a recovered journal to Shard does not happen during recovery

Thu, 11 Sep 2014 04:17:39 +0000

The RaftActor does not immediately apply state during recovery and instead relies on state changes or replication to apply the state. This can lead to a couple of issues. Firstly in a single node cluster the Leader will not reconstruct it's state till it gets the very first new transaction post recovery. Secondly this can lead to really slow application of the very first new transaction if the recovery process involves a lot of operations.

To reproduce the first problem do the following,

1. Start a single node cluster
2. Connect mininet to the cluster and create a small topology
3. Kill the controller (kill -9)
4. Exit mininet
5. Start the controller
6. Using restconf check if the nodes that were previously added are present

In this scenario the nodes should be present

OpenDaylight JIRA

[CONTROLLER-782] Clustering : Applying a recovered journal to Shard does not happen during recovery