Package org.apache.cassandra.spark.bulkwriter.token

Class MultiClusterReplicaAwareFailureHandler<I extends org.apache.cassandra.spark.common.model.CassandraInstance>

java.lang.Object

org.apache.cassandra.spark.bulkwriter.token.ReplicaAwareFailureHandler<I>

org.apache.cassandra.spark.bulkwriter.token.MultiClusterReplicaAwareFailureHandler<I>

Type Parameters:

I - CassandraInstance type

public class MultiClusterReplicaAwareFailureHandler<I extends org.apache.cassandra.spark.common.model.CassandraInstance>
extends ReplicaAwareFailureHandler<I>

A ReplicaAwareFailureHandler that can handle multiple clusters, including the case of single cluster.

Nested Class Summary

Nested classes/interfaces inherited from class org.apache.cassandra.spark.bulkwriter.token.ReplicaAwareFailureHandler

ReplicaAwareFailureHandler.ConsistencyFailurePerRange, ReplicaAwareFailureHandler.FailuresPerInstance

Constructor Summary

Constructors

Constructor	Description
MultiClusterReplicaAwareFailureHandler(org.apache.cassandra.spark.data.partitioner.Partitioner partitioner)

Method Summary

Modifier and Type	Method	Description
void	addFailure(com.google.common.collect.Range<java.math.BigInteger> tokenRange, I instance, java.lang.String errMessage)	Adds a new token range as a failed token range, with errors on given instance.
java.util.Set<I>	getFailedInstances()
java.util.List<ReplicaAwareFailureHandler.ConsistencyFailurePerRange>	getFailedRanges(TokenRangeMapping<I> tokenRangeMapping, JobInfo job, ClusterInfo cluster)	Given the number of failed instances for each token range, validates if the consistency guarantees are maintained for the job
protected java.util.List<ReplicaAwareFailureHandler.ConsistencyFailurePerRange>	getFailedRangesInternal(TokenRangeMapping<I> tokenRangeMapping, ConsistencyLevel cl, java.lang.String localDC, org.apache.cassandra.spark.data.ReplicationFactor replicationFactor)	Given the number of failed instances for each token range, validates if the consistency guarantees are maintained for the size of the ring and the consistency level.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

MultiClusterReplicaAwareFailureHandler

public MultiClusterReplicaAwareFailureHandler(org.apache.cassandra.spark.data.partitioner.Partitioner partitioner)

Method Detail

addFailure

public void addFailure(com.google.common.collect.Range<java.math.BigInteger> tokenRange,
                       I instance,
                       java.lang.String errMessage)

Description copied from class: ReplicaAwareFailureHandler

Adds a new token range as a failed token range, with errors on given instance.

It's guaranteed that failedRangesMap has overlapping ranges for the range we are trying to insert (Check constructor, we are adding complete ring first).

So the scheme is to get list of overlapping ranges first. For each overlapping range get the failure map. Make a copy of the map and add new failure to this map. It's important we make the copy and not use the one returned from failedRangesMap map. As our range could be overlapping partially and the map could be used by other range.

Specified by:

addFailure in class ReplicaAwareFailureHandler<I extends org.apache.cassandra.spark.common.model.CassandraInstance>

Parameters:

tokenRange - the range which failed

instance - the instance on which the range failed

errMessage - the error that occurred for this particular range/instance pair

getFailedInstances

public java.util.Set<I> getFailedInstances()

Specified by:

getFailedInstances in class ReplicaAwareFailureHandler<I extends org.apache.cassandra.spark.common.model.CassandraInstance>

Returns:

the set of all failed instances

getFailedRanges

public java.util.List<ReplicaAwareFailureHandler.ConsistencyFailurePerRange> getFailedRanges(TokenRangeMapping<I> tokenRangeMapping,
                                                                                             JobInfo job,
                                                                                             ClusterInfo cluster)

Description copied from class: ReplicaAwareFailureHandler

Given the number of failed instances for each token range, validates if the consistency guarantees are maintained for the job

Specified by:

getFailedRanges in class ReplicaAwareFailureHandler<I extends org.apache.cassandra.spark.common.model.CassandraInstance>

Parameters:

tokenRangeMapping - the mapping of token ranges to a Cassandra instance

job - the job to verify

cluster - cluster info

Returns:

list of failed token ranges that break consistency. This should ideally be empty for a successful operation.

getFailedRangesInternal

protected java.util.List<ReplicaAwareFailureHandler.ConsistencyFailurePerRange> getFailedRangesInternal(TokenRangeMapping<I> tokenRangeMapping,
                                                                                                        ConsistencyLevel cl,
                                                                                                        @Nullable
                                                                                                        java.lang.String localDC,
                                                                                                        org.apache.cassandra.spark.data.ReplicationFactor replicationFactor)

Description copied from class: ReplicaAwareFailureHandler

Given the number of failed instances for each token range, validates if the consistency guarantees are maintained for the size of the ring and the consistency level.

Specified by:

getFailedRangesInternal in class ReplicaAwareFailureHandler<I extends org.apache.cassandra.spark.common.model.CassandraInstance>

Parameters:

tokenRangeMapping - the mapping of token ranges to a Cassandra instance

cl - the desired consistency level

localDC - the local datacenter

replicationFactor - replication of the enclosing keyspace

Returns:

list of failed token ranges that break consistency. This should ideally be empty for a successful operation.