Method ShuffleShard

ring/partition_ring.go:193–206

ShuffleShard returns a subring for the provided identifier (eg. a tenant ID) and size (number of partitions). The algorithm used to build the subring is a shuffle sharder based on probabilistic hashing. We pick N unique partitions, walking the ring starting from random but predictable numbers. The

(identifier string, size int)

//   - Shuffling: probabilistically, for a large enough cluster each identifier gets a different
//     set of instances, with a reduced number of overlapping instances between two identifiers.
func (r *PartitionRing) ShuffleShard(identifier string, size int) (*PartitionRing, error) {
	if cached := r.shuffleShardCache.getSubring(identifier, size); cached != nil {
		return cached, nil
	}

	// No need to pass the time if there's no lookback.
	subring, err := r.shuffleShard(identifier, size, 0, time.Time{})
	if err != nil {
		return nil, err
	}

	r.shuffleShardCache.setSubring(identifier, size, subring)
	return subring, nil
}

// ShuffleShardWithLookback is like ShuffleShard() but the returned subring includes all instances
// that have been part of the identifier's shard in [now - lookbackPeriod, now] time window.