1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
|
/*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.spark.sql.kafka010
import java.{util => ju}
import scala.collection.mutable.ArrayBuffer
import org.apache.kafka.clients.consumer.ConsumerRecord
import org.apache.kafka.common.TopicPartition
import org.apache.spark.{Partition, SparkContext, TaskContext}
import org.apache.spark.partial.{BoundedDouble, PartialResult}
import org.apache.spark.rdd.RDD
import org.apache.spark.storage.StorageLevel
/** Offset range that one partition of the KafkaSourceRDD has to read */
private[kafka010] case class KafkaSourceRDDOffsetRange(
topicPartition: TopicPartition,
fromOffset: Long,
untilOffset: Long,
preferredLoc: Option[String]) {
def topic: String = topicPartition.topic
def partition: Int = topicPartition.partition
def size: Long = untilOffset - fromOffset
}
/** Partition of the KafkaSourceRDD */
private[kafka010] case class KafkaSourceRDDPartition(
index: Int, offsetRange: KafkaSourceRDDOffsetRange) extends Partition
/**
* An RDD that reads data from Kafka based on offset ranges across multiple partitions.
* Additionally, it allows preferred locations to be set for each topic + partition, so that
* the [[KafkaSource]] can ensure the same executor always reads the same topic + partition
* and cached KafkaConsuemrs (see [[CachedKafkaConsumer]] can be used read data efficiently.
*
* @param sc the [[SparkContext]]
* @param executorKafkaParams Kafka configuration for creating KafkaConsumer on the executors
* @param offsetRanges Offset ranges that define the Kafka data belonging to this RDD
*/
private[kafka010] class KafkaSourceRDD(
sc: SparkContext,
executorKafkaParams: ju.Map[String, Object],
offsetRanges: Seq[KafkaSourceRDDOffsetRange],
pollTimeoutMs: Long)
extends RDD[ConsumerRecord[Array[Byte], Array[Byte]]](sc, Nil) {
override def persist(newLevel: StorageLevel): this.type = {
logError("Kafka ConsumerRecord is not serializable. " +
"Use .map to extract fields before calling .persist or .window")
super.persist(newLevel)
}
override def getPartitions: Array[Partition] = {
offsetRanges.zipWithIndex.map { case (o, i) => new KafkaSourceRDDPartition(i, o) }.toArray
}
override def count(): Long = offsetRanges.map(_.size).sum
override def countApprox(timeout: Long, confidence: Double): PartialResult[BoundedDouble] = {
val c = count
new PartialResult(new BoundedDouble(c, 1.0, c, c), true)
}
override def isEmpty(): Boolean = count == 0L
override def take(num: Int): Array[ConsumerRecord[Array[Byte], Array[Byte]]] = {
val nonEmptyPartitions =
this.partitions.map(_.asInstanceOf[KafkaSourceRDDPartition]).filter(_.offsetRange.size > 0)
if (num < 1 || nonEmptyPartitions.isEmpty) {
return new Array[ConsumerRecord[Array[Byte], Array[Byte]]](0)
}
// Determine in advance how many messages need to be taken from each partition
val parts = nonEmptyPartitions.foldLeft(Map[Int, Int]()) { (result, part) =>
val remain = num - result.values.sum
if (remain > 0) {
val taken = Math.min(remain, part.offsetRange.size)
result + (part.index -> taken.toInt)
} else {
result
}
}
val buf = new ArrayBuffer[ConsumerRecord[Array[Byte], Array[Byte]]]
val res = context.runJob(
this,
(tc: TaskContext, it: Iterator[ConsumerRecord[Array[Byte], Array[Byte]]]) =>
it.take(parts(tc.partitionId)).toArray, parts.keys.toArray
)
res.foreach(buf ++= _)
buf.toArray
}
override def getPreferredLocations(split: Partition): Seq[String] = {
val part = split.asInstanceOf[KafkaSourceRDDPartition]
part.offsetRange.preferredLoc.map(Seq(_)).getOrElse(Seq.empty)
}
override def compute(
thePart: Partition,
context: TaskContext): Iterator[ConsumerRecord[Array[Byte], Array[Byte]]] = {
val range = thePart.asInstanceOf[KafkaSourceRDDPartition].offsetRange
assert(
range.fromOffset <= range.untilOffset,
s"Beginning offset ${range.fromOffset} is after the ending offset ${range.untilOffset} " +
s"for topic ${range.topic} partition ${range.partition}. " +
"You either provided an invalid fromOffset, or the Kafka topic has been damaged")
if (range.fromOffset == range.untilOffset) {
logInfo(s"Beginning offset ${range.fromOffset} is the same as ending offset " +
s"skipping ${range.topic} ${range.partition}")
Iterator.empty
} else {
val consumer = CachedKafkaConsumer.getOrCreate(
range.topic, range.partition, executorKafkaParams)
var requestOffset = range.fromOffset
logDebug(s"Creating iterator for $range")
new Iterator[ConsumerRecord[Array[Byte], Array[Byte]]]() {
override def hasNext(): Boolean = requestOffset < range.untilOffset
override def next(): ConsumerRecord[Array[Byte], Array[Byte]] = {
assert(hasNext(), "Can't call next() once untilOffset has been reached")
val r = consumer.get(requestOffset, pollTimeoutMs)
requestOffset += 1
r
}
}
}
}
}
|
