hash result changed a lot after bucket sized changed about go-jump-consistent-hash HOT 1 CLOSED

I'm not the paper author, I just ported the reference C++ implementation to Go. So I can't say much about the algorithm itself. I'm gonna have to to refer you to the paper¹ for details.

int32_t JumpConsistentHash(uint64_t key, int32_t num_buckets) {
  int64_t b = ­1, j = 0;
  while (j < num_buckets) {
    b=j;
    key = key * 2862933555777941757ULL + 1;
    j = (b + 1) * (double(1LL << 31) / double((key >> 33) + 1));
  }
  return b;
}

With that said. There's a "revamped" version called doublejump² that supposedly fixes some of the shortcomings of the original implementation (e.g. not being able to remove nodes). However I have no personal experience from using it myself.

from go-jump-consistent-hash.