I have a simple Ignite cluster with two nodes. It is a Java API based in-memory cluster and only uses IgniteCaches that are REPLICATED and SYNC, nothing fancy.
I'm trying to see how I can handle a potential split-brain case when there is a long gc or network hiccup. Luckily the specific cache data can be potentially thrown away, if needed, in order for a node to rejoin the cluster, so there is no issue with divergent state here. What I'm looking for is for the Ignite API that will trigger an Ignite node restart when communication between nodes is re-established. Note that I'm using the Java API so no control.sh scripts are available.
The solution seems to be relatively simple i.e. when a segmentation occurs, all nodes try to re-establish communication with one another and when they succeed all nodes except from the coordinator node restart in order to join the coordinator cluster. If the coordinator is down a new coordinator is elected/automatically chosen. Is there an API or interfaces I can implement to handle this case?
Similar products like NCache or Hazelcast seem to provide facilities to handle this so I'm guessing Apache Ignite has something similar.