Coarse parallel processing assigns one work item from a queue to each pod. After processing that item, the pod terminates, producing roughly one pod per work item. section 18.1.5
The queue and its worker logic belong in the application containers; Kubernetes runs those containers in one or more Job pods. Compared with fine parallel processing, coarse processing uses short-lived pods that handle single items, while fine processing lets a smaller group of pods handle multiple items. section 18.1.5
Processing queued months
A MongoDB collection stores twelve monthly work items. A Kubernetes Job can set completions to 6 and parallelism to 3; each worker script removes one item with findOneAndDelete(), processes it, prints completion, and exits. section 18.1.5
Figure 18.10 The difference between coarse and fine parallel processing
Compare how coarse processing creates one terminating pod per item with fine processing, where a smaller pod group drains the queue. section 18.1.5
