Hi everyone, i am using infinispan 5.3.0 and i am seeing a few things regarding the AsyncStore implementation:
1- If the store fails then it retries 3 times (which is not configurable) and then gives up, so it means, it does not store the object. Wouldnt it better to requeue the object in the write-behind queue in order to retry again later? (or do this behavior configurable?)
2- If the store fails and a rehashing happens due to a node joining or leaving then the store for the objects that were moved is not retry on the new node after the rehashing. I think we should have a flag on the entries that will tell us if the entry was stored or not and we can change it after store is sucessful, so when rehashing we will try only if the flag says the object is not stored so we do not miss stores. What do you think?
3- I am seeing the store operation is executed only for the primary copy of an object, which makes sense, but i think we should update the flag mentioned in point 2 after store sucessful and then update the same flag on the backups so we keep the flag in sync.