Affinity in the cluster is coupled to the topology.  Any affinity based
processing requires a stable topology.  Does Ignite guarantee that all nodes
are executing a transaction on the same affinity version?  Please correct me
if I'm wrong but based on my observation in  Race condition with partition
affinity mapping
there seems to be a gap here.  

Also, we would certainly benefit from 
1. exposing access to affintyTopologyVersion from Affinity interface for
tracking purposes and 
2. having functionality to await for specific affinity version before
proceeding with execution

Appreciate your thoughts

View this message in context:
Sent from the Apache Ignite Users mailing list archive at

Reply via email to