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
|
/*
* 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.execution.datasources
import java.util.concurrent.atomic.AtomicBoolean
import scala.collection.JavaConverters._
import com.google.common.cache._
import org.apache.hadoop.fs.{FileStatus, Path}
import org.apache.spark.internal.Logging
import org.apache.spark.sql.SparkSession
import org.apache.spark.util.SizeEstimator
/**
* Use [[FileStatusCache.getOrCreate()]] to construct a globally shared file status cache.
*/
object FileStatusCache {
private var sharedCache: SharedInMemoryCache = _
/**
* @return a new FileStatusCache based on session configuration. Cache memory quota is
* shared across all clients.
*/
def getOrCreate(session: SparkSession): FileStatusCache = synchronized {
if (session.sqlContext.conf.manageFilesourcePartitions &&
session.sqlContext.conf.filesourcePartitionFileCacheSize > 0) {
if (sharedCache == null) {
sharedCache = new SharedInMemoryCache(
session.sqlContext.conf.filesourcePartitionFileCacheSize)
}
sharedCache.createForNewClient()
} else {
NoopCache
}
}
def resetForTesting(): Unit = synchronized {
sharedCache = null
}
}
/**
* A cache of the leaf files of partition directories. We cache these files in order to speed
* up iterated queries over the same set of partitions. Otherwise, each query would have to
* hit remote storage in order to gather file statistics for physical planning.
*
* Each resolved catalog table has its own FileStatusCache. When the backing relation for the
* table is refreshed via refreshTable() or refreshByPath(), this cache will be invalidated.
*/
abstract class FileStatusCache {
/**
* @return the leaf files for the specified path from this cache, or None if not cached.
*/
def getLeafFiles(path: Path): Option[Array[FileStatus]] = None
/**
* Saves the given set of leaf files for a path in this cache.
*/
def putLeafFiles(path: Path, leafFiles: Array[FileStatus]): Unit
/**
* Invalidates all data held by this cache.
*/
def invalidateAll(): Unit
}
/**
* An implementation that caches partition file statuses in memory.
*
* @param maxSizeInBytes max allowable cache size before entries start getting evicted
*/
private class SharedInMemoryCache(maxSizeInBytes: Long) extends Logging {
// Opaque object that uniquely identifies a shared cache user
private type ClientId = Object
private val warnedAboutEviction = new AtomicBoolean(false)
// we use a composite cache key in order to distinguish entries inserted by different clients
private val cache: Cache[(ClientId, Path), Array[FileStatus]] = CacheBuilder.newBuilder()
.weigher(new Weigher[(ClientId, Path), Array[FileStatus]] {
override def weigh(key: (ClientId, Path), value: Array[FileStatus]): Int = {
(SizeEstimator.estimate(key) + SizeEstimator.estimate(value)).toInt
}})
.removalListener(new RemovalListener[(ClientId, Path), Array[FileStatus]]() {
override def onRemoval(removed: RemovalNotification[(ClientId, Path), Array[FileStatus]])
: Unit = {
if (removed.getCause == RemovalCause.SIZE &&
warnedAboutEviction.compareAndSet(false, true)) {
logWarning(
"Evicting cached table partition metadata from memory due to size constraints " +
"(spark.sql.hive.filesourcePartitionFileCacheSize = " + maxSizeInBytes + " bytes). " +
"This may impact query planning performance.")
}
}})
.maximumWeight(maxSizeInBytes)
.build[(ClientId, Path), Array[FileStatus]]()
/**
* @return a FileStatusCache that does not share any entries with any other client, but does
* share memory resources for the purpose of cache eviction.
*/
def createForNewClient(): FileStatusCache = new FileStatusCache {
val clientId = new Object()
override def getLeafFiles(path: Path): Option[Array[FileStatus]] = {
Option(cache.getIfPresent((clientId, path)))
}
override def putLeafFiles(path: Path, leafFiles: Array[FileStatus]): Unit = {
cache.put((clientId, path), leafFiles)
}
override def invalidateAll(): Unit = {
cache.asMap.asScala.foreach { case (key, value) =>
if (key._1 == clientId) {
cache.invalidate(key)
}
}
}
}
}
/**
* A non-caching implementation used when partition file status caching is disabled.
*/
object NoopCache extends FileStatusCache {
override def getLeafFiles(path: Path): Option[Array[FileStatus]] = None
override def putLeafFiles(path: Path, leafFiles: Array[FileStatus]): Unit = {}
override def invalidateAll(): Unit = {}
}
|
