org.apache.druid.server.coordinator.rules

Interface Rule

All Known Implementing Classes:

BroadcastDistributionRule, DropRule, ForeverBroadcastDistributionRule, ForeverDropRule, ForeverLoadRule, IntervalBroadcastDistributionRule, IntervalDropRule, IntervalLoadRule, LoadRule, PeriodBroadcastDistributionRule, PeriodDropBeforeRule, PeriodDropRule, PeriodLoadRule

public interface Rule

Method Summary

Modifier and Type	Method and Description
boolean	appliesTo(DataSegment segment, org.joda.time.DateTime referenceTimestamp)
boolean	appliesTo(org.joda.time.Interval interval, org.joda.time.DateTime referenceTimestamp)
boolean	canLoadSegments() Return true if this Rule can load segment onto one or more type of Druid node, otherwise return false.
String	getType()
CoordinatorStats	run(DruidCoordinator coordinator, DruidCoordinatorRuntimeParams params, DataSegment segment) DruidCoordinatorRuntimeParams.getUsedSegments() must not be called in Rule's code, because the used segments are not specified for the DruidCoordinatorRuntimeParams passed into Rule's code.
default void	updateUnderReplicated(Map<String,it.unimi.dsi.fastutil.objects.Object2LongMap<String>> underReplicatedPerTier, SegmentReplicantLookup segmentReplicantLookup, DataSegment segment) This method should update the with the replication count of the .
default void	updateUnderReplicatedWithClusterView(Map<String,it.unimi.dsi.fastutil.objects.Object2LongMap<String>> underReplicatedPerTier, SegmentReplicantLookup segmentReplicantLookup, DruidCluster cluster, DataSegment segment) This method should update the with the replication count of the taking into consideration the number of servers available in cluster that the segment can be replicated on.

Method Detail

getType

String getType()

appliesTo

boolean appliesTo(DataSegment segment,
                  org.joda.time.DateTime referenceTimestamp)

appliesTo

boolean appliesTo(org.joda.time.Interval interval,
                  org.joda.time.DateTime referenceTimestamp)

canLoadSegments

boolean canLoadSegments()

Return true if this Rule can load segment onto one or more type of Druid node, otherwise return false. Any Rule that returns true for this method should implement logic for calculating segment under replicated in updateUnderReplicated(java.util.Map<java.lang.String, it.unimi.dsi.fastutil.objects.Object2LongMap<java.lang.String>>, org.apache.druid.server.coordinator.SegmentReplicantLookup, org.apache.druid.timeline.DataSegment)

updateUnderReplicated

default void updateUnderReplicated(Map<String,it.unimi.dsi.fastutil.objects.Object2LongMap<String>> underReplicatedPerTier,
                                   SegmentReplicantLookup segmentReplicantLookup,
                                   DataSegment segment)

This method should update the with the replication count of the . Rule that returns true for canLoadSegments() must override this method. Note that is a map of tier -> { dataSource -> underReplicationCount }

updateUnderReplicatedWithClusterView

default void updateUnderReplicatedWithClusterView(Map<String,it.unimi.dsi.fastutil.objects.Object2LongMap<String>> underReplicatedPerTier,
                                                  SegmentReplicantLookup segmentReplicantLookup,
                                                  DruidCluster cluster,
                                                  DataSegment segment)

This method should update the with the replication count of the taking into consideration the number of servers available in cluster that the segment can be replicated on. Rule that returns true for canLoadSegments() must override this method. Note that is a map of tier -> { dataSource -> underReplicationCount }

run

CoordinatorStats run(DruidCoordinator coordinator,
                     DruidCoordinatorRuntimeParams params,
                     DataSegment segment)

DruidCoordinatorRuntimeParams.getUsedSegments() must not be called in Rule's code, because the used segments are not specified for the DruidCoordinatorRuntimeParams passed into Rule's code. This is because DruidCoordinatorRuntimeParams entangles two slightly different (nonexistent yet) abstractions: "CoordinatorDutyParams" and "RuleParams" which contain params that only CoordinatorDuty objects and Rules need, respectively. For example, ReplicationThrottler needs to belong only to "RuleParams", but not to "CoordinatorDutyParams". The opposite for the collection of used segments and DataSourcesSnapshot. See https://github.com/apache/druid/issues/7228