I want to understand the logic of how resolver spllits, it seems strange to me that the metric here is just
range begin key size plus an offset. So for a range written by a lot of short keys, and a range written by a lot of long keys, even though keys written per second might be the same, FDB will still consider the latter resolver is more heavily loaded, and thus will try to load balance between the two. anyone knows why FDB choose range begin key size as the metric?
1 Like