Distributed Spring Batch partitioning already needs a shared, persistent job repository. This article makes that database the whole coordination plane and drops the message broker, sizing work to the live node count, a per-job master with no standing coordinator, compare-and-set claiming that prevents double-execution, two-phase heartbeat failover, and where a broker is still the right choice.
By Janardhan Reddy Chejarla
InfoQ – Java
Article Categories:
Java