I’ve got a question about the range assignor RangeAssignor (kafka 3.3.2 API) that I understand is the default now. Using the example from the doc, let’s say that there are two topics t0 and t1 each having 3 partitions p0, p1, p2 give us t0p0, t0p1, t0p2, t1p0, t1p1, and t1p2. Then let’s say that there are 4 consumers divided into 2 consumer groups cg0 (c0, c1) and cg1 (c2, c3). Each consumer group just consumes from one topic so cg0 consumes t0 and cg1 consumes t1. Using the range assignor will result in
c0: [t0p0, t0p1]
c1: [t0p2]
c2: [t1p0, t1p1]
c3: [t1p2]
Let’s say that c0 and c2 are both running on one machine and c1, and c3 are on another. This will result in an uneven load on the machines. Does Kafka have some way of noticing this so it can e.g. give t1p1 to c3 instead?