| c1 | c2 | c3 | |
| range strategy | p0、p1 | p2、p3 | p4 |
| roundrobin strategy | p0、p3 | p1、p4 | p2 |
You may think that the two strategies are similar. Both strategies are: two partitions in c1, two partitions in c2, and one partition in c3
Let's look at the situation of multiple TOPICs:
| c1 | c2 | c3 | |
| range strategy | p1_0、p1_1、p2_0、p2_1 | p1_2、p1_3、p2_20、p2_3 | p1_4、p2_4 |
| roundrobin strategy | p1_0、p1_3、p2_1、p2_4 | p1_1、p1_4、p2_2 | p1_2、p2_0、p2_3 |
Under multiple TOPICs, the two strategies are different.
Under the range strategy: c1 four partitions, c2 four partitions, c3 two partitions
Under the roundrobin strategy: four partitions in c1, three partitions in c2, and three partitions in c3
Range strategy: For multiple topics, divide them by partition and range, and there may be a large difference in the number of partitions between consumers. (default)
roundrobin strategy: Multiple topics are taken as a whole, sorted according to topic + partition, and then distributed to consumers. Misassignments may occur in extreme cases. So kafka's default consumer partition strategy is range.