Package org.apache.druid.segment.loading

Interface StorageLocationSelectorStrategy

All Known Implementing Classes:

LeastBytesUsedStorageLocationSelectorStrategy, MostAvailableSizeStorageLocationSelectorStrategy, RandomStorageLocationSelectorStrategy, RoundRobinStorageLocationSelectorStrategy

public interface StorageLocationSelectorStrategy

This interface describes the storage location selection strategy which is responsible for ordering the available multiple StorageLocations for segment distribution. Only a snapshot of the locations is returned here. The implemntations currently do not handle all kinds of concurrency issues and accesses to the underlying storage. Please see https://github.com/apache/druid/pull/8038#discussion_r325520829 of PR https://github .com/apache/druid/pull/8038 for more details.

Method Summary

Modifier and Type	Method	Description
Iterator<StorageLocation>	getLocations()	Finds the best ordering of the StorageLocations to load a DataSegment according to the location selector strategy.

Method Details

getLocations

Iterator<StorageLocation> getLocations()

Finds the best ordering of the StorageLocations to load a DataSegment according to the location selector strategy. This method returns an iterator instead of a single best location. The caller is responsible for iterating over the locations and calling StorageLocation.reserve(java.lang.String, org.apache.druid.timeline.DataSegment) method. This is because a single location may be problematic like failed disk or might become unwritable for whatever reasons. This method can be called by different threads and so should be thread-safe.

Returns:

An iterator of StorageLocations from which the callers can iterate and pick a location.