diff --git common/src/java/org/apache/hadoop/hive/conf/HiveConf.java common/src/java/org/apache/hadoop/hive/conf/HiveConf.java index 43c53fc..73f5f24 100644 --- common/src/java/org/apache/hadoop/hive/conf/HiveConf.java +++ common/src/java/org/apache/hadoop/hive/conf/HiveConf.java @@ -2059,6 +2059,11 @@ public void setSparkConfigUpdated(boolean isSparkConfigUpdated) { "Channel logging level for remote Spark driver. One of {DEBUG, ERROR, INFO, TRACE, WARN}."), SPARK_RPC_SASL_MECHANISM("hive.spark.client.rpc.sasl.mechanisms", "DIGEST-MD5", "Name of the SASL mechanism to use for authentication."), + SPARK_DYNAMIC_RDD_CACHING("hive.spark.dynamic.rdd.caching", true, + "When dynamic rdd caching is enabled, Hive would find all the possible shared rdd, cache and reuse it in the" + + "generated Spark job finally."), + SPARK_DYNAMIC_RDD_CACHING_THRESHOLD("hive.spark.dynamic.rdd.caching.threshold", 100 * 1024 * 1024L, + "Maximum table rawDataSize which Hive may try to cache dynamically, default is 100MB."), NWAYJOINREORDER("hive.reorder.nway.joins", true, "Runs reordering of tables within single n-way join (i.e.: picks streamtable)"), HIVE_LOG_N_RECORDS("hive.log.every.n.records", 0L, new RangeValidator(0L, null), diff --git ql/src/java/org/apache/hadoop/hive/ql/exec/spark/CacheTran.java ql/src/java/org/apache/hadoop/hive/ql/exec/spark/CacheTran.java new file mode 100644 index 0000000..f3a53cf --- /dev/null +++ ql/src/java/org/apache/hadoop/hive/ql/exec/spark/CacheTran.java @@ -0,0 +1,54 @@ +/** + * 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.hadoop.hive.ql.exec.spark; + +import org.apache.hadoop.io.WritableComparable; +import org.apache.spark.api.java.JavaPairRDD; +import org.apache.spark.storage.StorageLevel; + +public abstract class CacheTran + implements SparkTran { + // whether to cache current RDD. + private boolean caching = false; + private JavaPairRDD cachedRDD; + + protected CacheTran(boolean cache) { + this.caching = cache; + } + + @Override + public JavaPairRDD transform( + JavaPairRDD input) { + if (caching) { + if (cachedRDD == null) { + cachedRDD = doTransform(input); + cachedRDD.persist(StorageLevel.MEMORY_ONLY()); + } + return cachedRDD; + } else { + return doTransform(input); + } + } + + public Boolean isCacheEnable() { + return caching; + } + + protected abstract JavaPairRDD doTransform(JavaPairRDD input); +} diff --git ql/src/java/org/apache/hadoop/hive/ql/exec/spark/LocalHiveSparkClient.java ql/src/java/org/apache/hadoop/hive/ql/exec/spark/LocalHiveSparkClient.java index 19d3fee..65d9ba7 100644 --- ql/src/java/org/apache/hadoop/hive/ql/exec/spark/LocalHiveSparkClient.java +++ ql/src/java/org/apache/hadoop/hive/ql/exec/spark/LocalHiveSparkClient.java @@ -136,7 +136,7 @@ public SparkJobRef execute(DriverContext driverContext, SparkWork sparkWork) thr // As we always use foreach action to submit RDD graph, it would only trigger one job. int jobId = future.jobIds().get(0); LocalSparkJobStatus sparkJobStatus = new LocalSparkJobStatus( - sc, jobId, jobMetricsListener, sparkCounters, plan.getCachedRDDIds(), future); + sc, jobId, jobMetricsListener, sparkCounters, future); return new LocalSparkJobRef(Integer.toString(jobId), hiveConf, sparkJobStatus, sc); } diff --git ql/src/java/org/apache/hadoop/hive/ql/exec/spark/MapInput.java ql/src/java/org/apache/hadoop/hive/ql/exec/spark/MapInput.java index 26cfebd..a7f9406 100644 --- ql/src/java/org/apache/hadoop/hive/ql/exec/spark/MapInput.java +++ ql/src/java/org/apache/hadoop/hive/ql/exec/spark/MapInput.java @@ -18,23 +18,15 @@ package org.apache.hadoop.hive.ql.exec.spark; -import org.apache.hadoop.conf.Configuration; +import com.google.common.base.Preconditions; import org.apache.hadoop.io.Writable; import org.apache.hadoop.io.WritableComparable; -import org.apache.hadoop.io.WritableUtils; import org.apache.spark.api.java.JavaPairRDD; -import org.apache.spark.api.java.function.PairFunction; -import org.apache.spark.storage.StorageLevel; - -import scala.Tuple2; - -import com.google.common.base.Preconditions; -public class MapInput implements SparkTran { private JavaPairRDD hadoopRDD; - private boolean toCache; private final SparkPlan sparkPlan; private String name = "MapInput"; @@ -44,47 +36,17 @@ public MapInput(SparkPlan sparkPlan, JavaPairRDD h public MapInput(SparkPlan sparkPlan, JavaPairRDD hadoopRDD, boolean toCache) { + super(toCache); this.hadoopRDD = hadoopRDD; - this.toCache = toCache; this.sparkPlan = sparkPlan; } - public void setToCache(boolean toCache) { - this.toCache = toCache; - } - @Override - public JavaPairRDD transform( + public JavaPairRDD doTransform( JavaPairRDD input) { Preconditions.checkArgument(input == null, "AssertionError: MapInput doesn't take any input"); - JavaPairRDD result; - if (toCache) { - result = hadoopRDD.mapToPair(new CopyFunction()); - sparkPlan.addCachedRDDId(result.id()); - result = result.persist(StorageLevel.MEMORY_AND_DISK()); - } else { - result = hadoopRDD; - } - return result; - } - - private static class CopyFunction implements PairFunction, - WritableComparable, Writable> { - - private transient Configuration conf; - - @Override - public Tuple2 - call(Tuple2 tuple) throws Exception { - if (conf == null) { - conf = new Configuration(); - } - - return new Tuple2(tuple._1(), - WritableUtils.clone(tuple._2(), conf)); - } - + return hadoopRDD; } @Override @@ -93,11 +55,6 @@ public String getName() { } @Override - public Boolean isCacheEnable() { - return new Boolean(toCache); - } - - @Override public void setName(String name) { this.name = name; } diff --git ql/src/java/org/apache/hadoop/hive/ql/exec/spark/MapTran.java ql/src/java/org/apache/hadoop/hive/ql/exec/spark/MapTran.java index 2170243..2a18991 100644 --- ql/src/java/org/apache/hadoop/hive/ql/exec/spark/MapTran.java +++ ql/src/java/org/apache/hadoop/hive/ql/exec/spark/MapTran.java @@ -22,12 +22,20 @@ import org.apache.hadoop.io.BytesWritable; import org.apache.spark.api.java.JavaPairRDD; -public class MapTran implements SparkTran { +public class MapTran extends CacheTran { private HiveMapFunction mapFunc; private String name = "MapTran"; + public MapTran() { + this(false); + } + + public MapTran(boolean cache) { + super(cache); + } + @Override - public JavaPairRDD transform( + public JavaPairRDD doTransform( JavaPairRDD input) { return input.mapPartitionsToPair(mapFunc); } @@ -42,11 +50,6 @@ public String getName() { } @Override - public Boolean isCacheEnable() { - return null; - } - - @Override public void setName(String name) { this.name = name; } diff --git ql/src/java/org/apache/hadoop/hive/ql/exec/spark/ReduceTran.java ql/src/java/org/apache/hadoop/hive/ql/exec/spark/ReduceTran.java index e60dfac..3d56876 100644 --- ql/src/java/org/apache/hadoop/hive/ql/exec/spark/ReduceTran.java +++ ql/src/java/org/apache/hadoop/hive/ql/exec/spark/ReduceTran.java @@ -22,12 +22,20 @@ import org.apache.hadoop.io.BytesWritable; import org.apache.spark.api.java.JavaPairRDD; -public class ReduceTran implements SparkTran, HiveKey, BytesWritable> { +public class ReduceTran extends CacheTran, HiveKey, BytesWritable> { private HiveReduceFunction reduceFunc; private String name = "Reduce"; + public ReduceTran() { + this(false); + } + + public ReduceTran(boolean caching) { + super(caching); + } + @Override - public JavaPairRDD transform( + public JavaPairRDD doTransform( JavaPairRDD> input) { return input.mapPartitionsToPair(reduceFunc); } @@ -42,11 +50,6 @@ public String getName() { } @Override - public Boolean isCacheEnable() { - return null; - } - - @Override public void setName(String name) { this.name = name; } diff --git ql/src/java/org/apache/hadoop/hive/ql/exec/spark/RemoteHiveSparkClient.java ql/src/java/org/apache/hadoop/hive/ql/exec/spark/RemoteHiveSparkClient.java index 8b15099..2a4686b 100644 --- ql/src/java/org/apache/hadoop/hive/ql/exec/spark/RemoteHiveSparkClient.java +++ ql/src/java/org/apache/hadoop/hive/ql/exec/spark/RemoteHiveSparkClient.java @@ -259,7 +259,7 @@ public Serializable call(JobContext jc) throws Exception { JavaPairRDD finalRDD = plan.generateGraph(); // We use Spark RDD async action to submit job as it's the only way to get jobId now. JavaFutureAction future = finalRDD.foreachAsync(HiveVoidFunction.getInstance()); - jc.monitor(future, sparkCounters, plan.getCachedRDDIds()); + jc.monitor(future, sparkCounters); return null; } diff --git ql/src/java/org/apache/hadoop/hive/ql/exec/spark/ShuffleTran.java ql/src/java/org/apache/hadoop/hive/ql/exec/spark/ShuffleTran.java index a774395..12e67e6 100644 --- ql/src/java/org/apache/hadoop/hive/ql/exec/spark/ShuffleTran.java +++ ql/src/java/org/apache/hadoop/hive/ql/exec/spark/ShuffleTran.java @@ -23,10 +23,9 @@ import org.apache.spark.api.java.JavaPairRDD; import org.apache.spark.storage.StorageLevel; -public class ShuffleTran implements SparkTran> { +public class ShuffleTran extends CacheTran> { private final SparkShuffler shuffler; private final int numOfPartitions; - private final boolean toCache; private final SparkPlan sparkPlan; private String name = "Shuffle"; @@ -35,19 +34,15 @@ public ShuffleTran(SparkPlan sparkPlan, SparkShuffler sf, int n) { } public ShuffleTran(SparkPlan sparkPlan, SparkShuffler sf, int n, boolean toCache) { + super(toCache); shuffler = sf; numOfPartitions = n; - this.toCache = toCache; this.sparkPlan = sparkPlan; } @Override - public JavaPairRDD> transform(JavaPairRDD input) { + public JavaPairRDD> doTransform(JavaPairRDD input) { JavaPairRDD> result = shuffler.shuffle(input, numOfPartitions); - if (toCache) { - sparkPlan.addCachedRDDId(result.id()); - result = result.persist(StorageLevel.MEMORY_AND_DISK()); - } return result; } @@ -61,11 +56,6 @@ public String getName() { } @Override - public Boolean isCacheEnable() { - return new Boolean(toCache); - } - - @Override public void setName(String name) { this.name = name; } diff --git ql/src/java/org/apache/hadoop/hive/ql/exec/spark/SparkPlan.java ql/src/java/org/apache/hadoop/hive/ql/exec/spark/SparkPlan.java index ee5c78a..0cb20c9 100644 --- ql/src/java/org/apache/hadoop/hive/ql/exec/spark/SparkPlan.java +++ ql/src/java/org/apache/hadoop/hive/ql/exec/spark/SparkPlan.java @@ -88,6 +88,7 @@ } perfLogger.PerfLogEnd(CLASS_NAME, PerfLogger.SPARK_BUILD_RDD_GRAPH); + LOG.info("print generated spark rdd graph:\n" + SparkUtilities.rddGraphToString(finalRDD)); return finalRDD; } @@ -238,14 +239,6 @@ public void addTran(SparkTran tran) { leafTrans.add(tran); } - public void addCachedRDDId(int rddId) { - cachedRDDIds.add(rddId); - } - - public Set getCachedRDDIds() { - return cachedRDDIds; - } - /** * This method returns a topologically sorted list of SparkTran. */ diff --git ql/src/java/org/apache/hadoop/hive/ql/exec/spark/SparkPlanGenerator.java ql/src/java/org/apache/hadoop/hive/ql/exec/spark/SparkPlanGenerator.java index 3f240f5..c91de08 100644 --- ql/src/java/org/apache/hadoop/hive/ql/exec/spark/SparkPlanGenerator.java +++ ql/src/java/org/apache/hadoop/hive/ql/exec/spark/SparkPlanGenerator.java @@ -22,9 +22,13 @@ import java.util.List; import java.util.Map; import java.util.Set; +import java.util.concurrent.ConcurrentMap; +import java.util.concurrent.atomic.AtomicInteger; import com.google.common.base.Preconditions; +import com.google.common.collect.Maps; +import org.apache.commons.lang3.tuple.Pair; import org.apache.commons.logging.Log; import org.apache.commons.logging.LogFactory; import org.apache.hadoop.fs.Path; @@ -69,12 +73,12 @@ private Context context; private Path scratchDir; private SparkReporter sparkReporter; - private Map cloneToWork; private final Map workToTranMap; - private final Map workToParentWorkTranMap; // a map from each BaseWork to its cloned JobConf private final Map workToJobConf; + private List cachingWorks; + public SparkPlanGenerator( JavaSparkContext sc, Context context, @@ -87,17 +91,15 @@ public SparkPlanGenerator( this.jobConf = jobConf; this.scratchDir = scratchDir; this.workToTranMap = new HashMap(); - this.workToParentWorkTranMap = new HashMap(); this.sparkReporter = sparkReporter; this.workToJobConf = new HashMap(); } public SparkPlan generate(SparkWork sparkWork) throws Exception { perfLogger.PerfLogBegin(CLASS_NAME, PerfLogger.SPARK_BUILD_PLAN); + cachingWorks = sparkWork.getCachingWorks(); SparkPlan sparkPlan = new SparkPlan(); - cloneToWork = sparkWork.getCloneToWork(); workToTranMap.clear(); - workToParentWorkTranMap.clear(); try { for (BaseWork work : sparkWork.getAllWork()) { @@ -122,13 +124,6 @@ public SparkPlan generate(SparkWork sparkWork) throws Exception { // Generate (possibly get from a cached result) parent SparkTran private SparkTran generateParentTran(SparkPlan sparkPlan, SparkWork sparkWork, BaseWork work) throws Exception { - if (cloneToWork.containsKey(work)) { - BaseWork originalWork = cloneToWork.get(work); - if (workToParentWorkTranMap.containsKey(originalWork)) { - return workToParentWorkTranMap.get(originalWork); - } - } - SparkTran result; if (work instanceof MapWork) { result = generateMapInput(sparkPlan, (MapWork)work); @@ -136,7 +131,7 @@ private SparkTran generateParentTran(SparkPlan sparkPlan, SparkWork sparkWork, } else if (work instanceof ReduceWork) { List parentWorks = sparkWork.getParents(work); result = generate(sparkPlan, - sparkWork.getEdgeProperty(parentWorks.get(0), work), cloneToWork.containsKey(work)); + sparkWork.getEdgeProperty(parentWorks.get(0), work)); sparkPlan.addTran(result); for (BaseWork parentWork : parentWorks) { sparkPlan.connect(workToTranMap.get(parentWork), result); @@ -146,10 +141,6 @@ private SparkTran generateParentTran(SparkPlan sparkPlan, SparkWork sparkWork, + "but found " + work.getClass().getName()); } - if (cloneToWork.containsKey(work)) { - workToParentWorkTranMap.put(cloneToWork.get(work), result); - } - return result; } @@ -181,18 +172,17 @@ private SparkTran generateParentTran(SparkPlan sparkPlan, SparkWork sparkWork, @SuppressWarnings("unchecked") private MapInput generateMapInput(SparkPlan sparkPlan, MapWork mapWork) - throws Exception { + throws Exception { JobConf jobConf = cloneJobConf(mapWork); Class ifClass = getInputFormat(jobConf, mapWork); JavaPairRDD hadoopRDD = sc.hadoopRDD(jobConf, ifClass, - WritableComparable.class, Writable.class); - // Caching is disabled for MapInput due to HIVE-8920 - MapInput result = new MapInput(sparkPlan, hadoopRDD, false/*cloneToWork.containsKey(mapWork)*/); + WritableComparable.class, Writable.class); + MapInput result = new MapInput(sparkPlan, hadoopRDD); return result; } - private ShuffleTran generate(SparkPlan sparkPlan, SparkEdgeProperty edge, boolean toCache) { + private ShuffleTran generate(SparkPlan sparkPlan, SparkEdgeProperty edge) { Preconditions.checkArgument(!edge.isShuffleNone(), "AssertionError: SHUFFLE_NONE should only be used for UnionWork."); SparkShuffler shuffler; @@ -203,7 +193,7 @@ private ShuffleTran generate(SparkPlan sparkPlan, SparkEdgeProperty edge, boolea } else { shuffler = new GroupByShuffler(); } - return new ShuffleTran(sparkPlan, shuffler, edge.getNumPartitions(), toCache); + return new ShuffleTran(sparkPlan, shuffler, edge.getNumPartitions()); } private SparkTran generate(BaseWork work) throws Exception { @@ -211,13 +201,17 @@ private SparkTran generate(BaseWork work) throws Exception { JobConf newJobConf = cloneJobConf(work); checkSpecs(work, newJobConf); byte[] confBytes = KryoSerializer.serializeJobConf(newJobConf); + boolean caching = false; + if (cachingWorks != null && cachingWorks.contains(work)) { + caching = true; + } if (work instanceof MapWork) { - MapTran mapTran = new MapTran(); + MapTran mapTran = new MapTran(caching); HiveMapFunction mapFunc = new HiveMapFunction(confBytes, sparkReporter); mapTran.setMapFunction(mapFunc); return mapTran; } else if (work instanceof ReduceWork) { - ReduceTran reduceTran = new ReduceTran(); + ReduceTran reduceTran = new ReduceTran(caching); HiveReduceFunction reduceFunc = new HiveReduceFunction(confBytes, sparkReporter); reduceTran.setReduceFunction(reduceFunc); return reduceTran; @@ -295,5 +289,4 @@ private void initStatsPublisher(BaseWork work) throws HiveException { } } } - } diff --git ql/src/java/org/apache/hadoop/hive/ql/exec/spark/SparkUtilities.java ql/src/java/org/apache/hadoop/hive/ql/exec/spark/SparkUtilities.java index e6c845c..2529d8a 100644 --- ql/src/java/org/apache/hadoop/hive/ql/exec/spark/SparkUtilities.java +++ ql/src/java/org/apache/hadoop/hive/ql/exec/spark/SparkUtilities.java @@ -21,6 +21,7 @@ import java.io.IOException; import java.net.URI; import java.net.URISyntaxException; +import java.util.Collection; import java.util.UUID; import org.apache.commons.io.FilenameUtils; @@ -34,7 +35,12 @@ import org.apache.hadoop.hive.ql.metadata.HiveException; import org.apache.hadoop.hive.ql.session.SessionState; import org.apache.hadoop.io.BytesWritable; +import org.apache.spark.Dependency; import org.apache.spark.SparkConf; +import org.apache.spark.api.java.JavaPairRDD; +import org.apache.spark.rdd.RDD; +import org.apache.spark.rdd.UnionRDD; +import scala.collection.JavaConversions; /** * Contains utilities methods used as part of Spark tasks. @@ -122,4 +128,33 @@ public static SparkSession getSparkSession(HiveConf conf, SessionState.get().setSparkSession(sparkSession); return sparkSession; } + + public static String rddGraphToString(JavaPairRDD rdd) { + StringBuilder sb = new StringBuilder(); + rddToString(rdd.rdd(), sb, ""); + return sb.toString(); + } + + private static void rddToString(RDD rdd, StringBuilder sb, String offset) { + sb.append(offset).append(rdd.getClass().getCanonicalName()).append("[").append(rdd.hashCode()).append("]"); + if (rdd.getStorageLevel().useMemory()) { + sb.append("(cached)"); + } + sb.append("\n"); + Collection dependencies = JavaConversions.asJavaCollection(rdd.dependencies()); + if (dependencies != null) { + offset += "\t"; + for (Dependency dependency : dependencies) { + RDD parentRdd = dependency.rdd(); + rddToString(parentRdd, sb, offset); + } + } else if (rdd instanceof UnionRDD) { + UnionRDD unionRDD = (UnionRDD) rdd; + offset += "\t"; + Collection parentRdds = JavaConversions.asJavaCollection(unionRDD.rdds()); + for (RDD parentRdd : parentRdds) { + rddToString(parentRdd, sb, offset); + } + } + } } diff --git ql/src/java/org/apache/hadoop/hive/ql/exec/spark/status/impl/LocalSparkJobStatus.java ql/src/java/org/apache/hadoop/hive/ql/exec/spark/status/impl/LocalSparkJobStatus.java index 5d62596..4081232 100644 --- ql/src/java/org/apache/hadoop/hive/ql/exec/spark/status/impl/LocalSparkJobStatus.java +++ ql/src/java/org/apache/hadoop/hive/ql/exec/spark/status/impl/LocalSparkJobStatus.java @@ -52,16 +52,14 @@ private JobMetricsListener jobMetricsListener; private SparkCounters sparkCounters; private JavaFutureAction future; - private Set cachedRDDIds; public LocalSparkJobStatus(JavaSparkContext sparkContext, int jobId, JobMetricsListener jobMetricsListener, SparkCounters sparkCounters, - Set cachedRDDIds, JavaFutureAction future) { + JavaFutureAction future) { this.sparkContext = sparkContext; this.jobId = jobId; this.jobMetricsListener = jobMetricsListener; this.sparkCounters = sparkCounters; - this.cachedRDDIds = cachedRDDIds; this.future = future; } @@ -141,11 +139,6 @@ public SparkStatistics getSparkStatistics() { @Override public void cleanup() { jobMetricsListener.cleanup(jobId); - if (cachedRDDIds != null) { - for (Integer cachedRDDId: cachedRDDIds) { - sparkContext.sc().unpersistRDD(cachedRDDId, false); - } - } } private Map combineJobLevelMetrics(Map> jobMetric) { diff --git ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/SparkMapJoinResolver.java ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/SparkMapJoinResolver.java index 8e56263..7b3026b 100644 --- ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/SparkMapJoinResolver.java +++ ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/SparkMapJoinResolver.java @@ -163,8 +163,6 @@ private void moveWork(SparkWork sparkWork, BaseWork work, SparkWork targetWork) // Create a new SparkWork for all the small tables of this work SparkWork parentWork = new SparkWork(physicalContext.conf.getVar(HiveConf.ConfVars.HIVEQUERYID)); - // copy cloneToWork to ensure RDD cache still works - parentWork.setCloneToWork(sparkWork.getCloneToWork()); dependencyGraph.get(targetWork).add(parentWork); dependencyGraph.put(parentWork, new ArrayList()); diff --git ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/SparkRddCachingResolver.java ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/SparkRddCachingResolver.java new file mode 100644 index 0000000..8ff8fd7 --- /dev/null +++ ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/SparkRddCachingResolver.java @@ -0,0 +1,181 @@ +/** + * 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.hadoop.hive.ql.optimizer.physical; + +import java.io.Serializable; +import java.util.ArrayList; +import java.util.HashMap; +import java.util.HashSet; +import java.util.Iterator; +import java.util.LinkedHashMap; +import java.util.LinkedList; +import java.util.List; +import java.util.Map; +import java.util.Set; +import java.util.Stack; +import java.util.UUID; + +import com.google.common.base.Preconditions; +import com.google.common.collect.Lists; +import com.google.common.collect.Maps; +import com.google.common.collect.Sets; +import org.apache.commons.lang3.tuple.MutablePair; +import org.apache.commons.lang3.tuple.Pair; +import org.apache.commons.logging.Log; +import org.apache.commons.logging.LogFactory; +import org.apache.hadoop.conf.Configuration; +import org.apache.hadoop.hive.conf.HiveConf; +import org.apache.hadoop.hive.ql.DriverContext; +import org.apache.hadoop.hive.ql.exec.MapOperator; +import org.apache.hadoop.hive.ql.exec.Operator; +import org.apache.hadoop.hive.ql.exec.TableScanOperator; +import org.apache.hadoop.hive.ql.exec.Task; +import org.apache.hadoop.hive.ql.exec.spark.SparkTask; +import org.apache.hadoop.hive.ql.io.orc.OrcInputFormat; +import org.apache.hadoop.hive.ql.io.orc.OrcNewInputFormat; +import org.apache.hadoop.hive.ql.lib.Dispatcher; +import org.apache.hadoop.hive.ql.lib.Node; +import org.apache.hadoop.hive.ql.lib.TaskGraphWalker; +import org.apache.hadoop.hive.ql.metadata.HiveException; +import org.apache.hadoop.hive.ql.metadata.Table; +import org.apache.hadoop.hive.ql.parse.SemanticException; +import org.apache.hadoop.hive.ql.plan.BaseWork; +import org.apache.hadoop.hive.ql.plan.MapWork; +import org.apache.hadoop.hive.ql.plan.OperatorDesc; +import org.apache.hadoop.hive.ql.plan.PartitionDesc; +import org.apache.hadoop.hive.ql.plan.SparkWork; +import org.apache.hadoop.hive.ql.plan.TableDesc; +import org.apache.hadoop.hive.ql.plan.TableScanDesc; +import org.apache.hadoop.hive.ql.plan.api.StageType; +import org.apache.hadoop.hive.ql.stats.StatsUtils; +import org.apache.hadoop.hive.serde2.Deserializer; +import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspector; +import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspectorConverters; +import org.apache.hadoop.hive.serde2.objectinspector.StructObjectInspector; +import org.apache.hadoop.mapred.InputFormat; +import parquet.hadoop.ParquetInputFormat; + +/** + * A Hive query may scan the same table multi times, or even share same part in the query, + * so cache the shared RDD would reduce IO and CPU cost, which help to improve Hive performance. + * SparkRddCachingResolver is in charge of walking through Tasks, and parttern matching all + * the cacheable MapWork/ReduceWork. + */ +public class SparkRddCachingResolver implements PhysicalPlanResolver { + + private static final Log LOG = LogFactory.getLog(SparkRddCachingResolver.class); + + @Override + public PhysicalContext resolve(PhysicalContext pctx) throws SemanticException { + List sparkTasks = getAllSparkTasks(pctx); + ArrayList topNodes = new ArrayList(); + if (sparkTasks.size() > 0) { + topNodes.addAll(sparkTasks); + TaskGraphWalker sparkWorkOgw = new TaskGraphWalker(new ShareParentWorkMatching(pctx.getConf())); + sparkWorkOgw.startWalking(topNodes, null); + } else { + LOG.info("No SparkWork found, skip SparkRddCachingResolver."); + } + + return pctx; + } + + /** + * A Work graph like: + * MapWork1 + * / \ + * ReduceWork1 ReduceWork2 + * would be translated into RDD graph like: + * MapPartitionsRDD1 + * / \ + * ShuffledRDD1 ShuffledRDD2 + * / \ + * MapPartitionsRDD3 MapPartitionsRDD4 + * In the Spark implementation, MapPartitionsRDD1 would be computed twice, so cache it may improve performance. + * ShareParentWorkMatching try to match all the works with multi children in SparkWork here. + */ + class ShareParentWorkMatching implements Dispatcher { + private HiveConf hiveConf; + + ShareParentWorkMatching(HiveConf hiveConf) { + this.hiveConf = hiveConf; + } + + @Override + public Object dispatch(Node nd, Stack stack, Object... nodeOutputs) throws SemanticException { + if (nd instanceof SparkTask) { + SparkTask sparkTask = (SparkTask) nd; + SparkWork sparkWork = sparkTask.getWork(); + List toCacheWorks = Lists.newLinkedList(); + List> pairs = Lists.newArrayList(sparkWork.getEdgeProperties().keySet()); + int size = pairs.size(); + for (int i = 0; i < size; i++) { + Pair first = pairs.get(i); + for (int j = i + 1; j < size; j++) { + Pair second = pairs.get(j); + if (first.getKey().equals(second.getKey()) && !first.getValue().equals(second.getValue())) { + BaseWork work = first.getKey(); + long estimatedDataSize = getWorkDataSize(work); + long threshold = HiveConf.getLongVar(hiveConf, HiveConf.ConfVars.SPARK_DYNAMIC_RDD_CACHING_THRESHOLD); + if (estimatedDataSize > threshold) { + continue; + } else { + toCacheWorks.add(work); + } + } + } + } + sparkWork.setCachingWorks(toCacheWorks); + } + return null; + } + } + + private List getAllSparkTasks(PhysicalContext pctx) { + List sparkTasks = new LinkedList(); + List> rootTasks = pctx.getRootTasks(); + if (rootTasks != null) { + for (Task task : rootTasks) { + getSparkTask(task, sparkTasks); + } + } + return sparkTasks; + } + + private void getSparkTask(Task task, List sparkTasks) { + if (task instanceof SparkTask) { + sparkTasks.add((SparkTask) task); + List childTasks = task.getChildTasks(); + if (childTasks != null) { + for (Task childTask : childTasks) { + getSparkTask(childTask, sparkTasks); + } + } + } + } + + private long getWorkDataSize(BaseWork work) { + long size = 0; + Set> leafOperators = work.getAllLeafOperators(); + for (Operator operator : leafOperators) { + size += operator.getStatistics().getDataSize(); + } + return size; + } +} diff --git ql/src/java/org/apache/hadoop/hive/ql/optimizer/spark/SparkSkewJoinProcFactory.java ql/src/java/org/apache/hadoop/hive/ql/optimizer/spark/SparkSkewJoinProcFactory.java index 5990d17..aba6954 100644 --- ql/src/java/org/apache/hadoop/hive/ql/optimizer/spark/SparkSkewJoinProcFactory.java +++ ql/src/java/org/apache/hadoop/hive/ql/optimizer/spark/SparkSkewJoinProcFactory.java @@ -121,7 +121,6 @@ private static void splitTask(SparkTask currentTask, ReduceWork reduceWork, // remove them from current spark work for (BaseWork baseWork : newWork.getAllWorkUnsorted()) { currentWork.remove(baseWork); - currentWork.getCloneToWork().remove(baseWork); } // create TS to read intermediate data Context baseCtx = parseContext.getContext(); diff --git ql/src/java/org/apache/hadoop/hive/ql/optimizer/spark/SplitSparkWorkResolver.java ql/src/java/org/apache/hadoop/hive/ql/optimizer/spark/SplitSparkWorkResolver.java index fb20080..67252da 100644 --- ql/src/java/org/apache/hadoop/hive/ql/optimizer/spark/SplitSparkWorkResolver.java +++ ql/src/java/org/apache/hadoop/hive/ql/optimizer/spark/SplitSparkWorkResolver.java @@ -128,7 +128,6 @@ private void splitBaseWork(SparkWork sparkWork, BaseWork parentWork, List> rootTasks, Pa LOG.debug("Skipping stage id rearranger"); } + if (conf.getBoolVar(HiveConf.ConfVars.SPARK_DYNAMIC_RDD_CACHING)) { + physicalCtx = new SparkRddCachingResolver().resolve(physicalCtx); + } + PERF_LOGGER.PerfLogEnd(CLASS_NAME, PerfLogger.SPARK_OPTIMIZE_TASK_TREE); return; } diff --git ql/src/java/org/apache/hadoop/hive/ql/plan/SparkWork.java ql/src/java/org/apache/hadoop/hive/ql/plan/SparkWork.java index bb5dd79..56d9714 100644 --- ql/src/java/org/apache/hadoop/hive/ql/plan/SparkWork.java +++ ql/src/java/org/apache/hadoop/hive/ql/plan/SparkWork.java @@ -19,23 +19,24 @@ package org.apache.hadoop.hive.ql.plan; import java.io.Serializable; - import java.util.ArrayList; import java.util.Collection; import java.util.Collections; import java.util.HashMap; import java.util.HashSet; +import java.util.LinkedHashMap; import java.util.LinkedHashSet; import java.util.LinkedList; -import java.util.LinkedHashMap; import java.util.List; import java.util.Map; import java.util.Set; +import com.google.common.base.Preconditions; +import com.google.common.collect.ImmutableMap; +import com.google.common.collect.Lists; import org.apache.commons.lang3.tuple.ImmutablePair; import org.apache.commons.lang3.tuple.Pair; -import com.google.common.base.Preconditions; /** * This class encapsulates all the work objects that can be executed @@ -60,11 +61,10 @@ private Map> requiredCounterPrefix; - private Map cloneToWork; + private List cachingWorks; public SparkWork(String name) { this.name = name + ":" + (++counter); - cloneToWork = new HashMap(); } @@ -416,11 +416,28 @@ public String toString() { return result; } - public Map getCloneToWork() { - return cloneToWork; + /** + * @return all map works of this spark work, in sorted order. + */ + public List getAllMapWork() { + List result = Lists.newLinkedList(); + for (BaseWork work : getAllWork()) { + if (work instanceof MapWork) { + result.add((MapWork) work); + } + } + return result; + } + + public Map, SparkEdgeProperty> getEdgeProperties() { + return ImmutableMap.copyOf(edgeProperties); + } + + public List getCachingWorks() { + return cachingWorks; } - public void setCloneToWork(Map cloneToWork) { - this.cloneToWork = cloneToWork; + public void setCachingWorks(List cachingWorks) { + this.cachingWorks = cachingWorks; } } diff --git spark-client/src/main/java/org/apache/hive/spark/client/JobContext.java spark-client/src/main/java/org/apache/hive/spark/client/JobContext.java index af6332e..590a604 100644 --- spark-client/src/main/java/org/apache/hive/spark/client/JobContext.java +++ spark-client/src/main/java/org/apache/hive/spark/client/JobContext.java @@ -47,7 +47,7 @@ * @return The job (unmodified). */ JavaFutureAction monitor( - JavaFutureAction job, SparkCounters sparkCounters, Set cachedRDDIds); + JavaFutureAction job, SparkCounters sparkCounters); /** * Return a map from client job Id to corresponding JavaFutureActions. diff --git spark-client/src/main/java/org/apache/hive/spark/client/JobContextImpl.java spark-client/src/main/java/org/apache/hive/spark/client/JobContextImpl.java index beed8a3..7732476 100644 --- spark-client/src/main/java/org/apache/hive/spark/client/JobContextImpl.java +++ spark-client/src/main/java/org/apache/hive/spark/client/JobContextImpl.java @@ -54,8 +54,8 @@ public JavaSparkContext sc() { @Override public JavaFutureAction monitor(JavaFutureAction job, - SparkCounters sparkCounters, Set cachedRDDIds) { - monitorCb.get().call(job, sparkCounters, cachedRDDIds); + SparkCounters sparkCounters) { + monitorCb.get().call(job, sparkCounters); return job; } diff --git spark-client/src/main/java/org/apache/hive/spark/client/MonitorCallback.java spark-client/src/main/java/org/apache/hive/spark/client/MonitorCallback.java index e1e899e..b09f28e 100644 --- spark-client/src/main/java/org/apache/hive/spark/client/MonitorCallback.java +++ spark-client/src/main/java/org/apache/hive/spark/client/MonitorCallback.java @@ -25,6 +25,6 @@ interface MonitorCallback { - void call(JavaFutureAction future, SparkCounters sparkCounters, Set cachedRDDIds); + void call(JavaFutureAction future, SparkCounters sparkCounters); } diff --git spark-client/src/main/java/org/apache/hive/spark/client/RemoteDriver.java spark-client/src/main/java/org/apache/hive/spark/client/RemoteDriver.java index b77c9e8..1703309 100644 --- spark-client/src/main/java/org/apache/hive/spark/client/RemoteDriver.java +++ spark-client/src/main/java/org/apache/hive/spark/client/RemoteDriver.java @@ -318,7 +318,7 @@ private Object handle(ChannelHandlerContext ctx, SyncJobRequest msg) throws Exce jc.setMonitorCb(new MonitorCallback() { @Override public void call(JavaFutureAction future, - SparkCounters sparkCounters, Set cachedRDDIds) { + SparkCounters sparkCounters) { throw new IllegalStateException( "JobContext.monitor() is not available for synchronous jobs."); } @@ -338,8 +338,6 @@ public void call(JavaFutureAction future, private final List> jobs; private final AtomicInteger completed; private SparkCounters sparkCounters; - private Set cachedRDDIds; - private Future future; JobWrapper(BaseProtocol.JobRequest req) { @@ -347,7 +345,6 @@ public void call(JavaFutureAction future, this.jobs = Lists.newArrayList(); this.completed = new AtomicInteger(); this.sparkCounters = null; - this.cachedRDDIds = null; } @Override @@ -358,8 +355,8 @@ public Void call() throws Exception { jc.setMonitorCb(new MonitorCallback() { @Override public void call(JavaFutureAction future, - SparkCounters sparkCounters, Set cachedRDDIds) { - monitorJob(future, sparkCounters, cachedRDDIds); + SparkCounters sparkCounters) { + monitorJob(future, sparkCounters); } }); @@ -393,7 +390,6 @@ public void call(JavaFutureAction future, } finally { jc.setMonitorCb(null); activeJobs.remove(req.id); - releaseCache(); } return null; } @@ -409,30 +405,14 @@ void jobDone() { } } - /** - * Release cached RDDs as soon as the job is done. - * This is different from local Spark client so as - * to save a RPC call/trip, avoid passing cached RDD - * id information around. Otherwise, we can follow - * the local Spark client way to be consistent. - */ - void releaseCache() { - if (cachedRDDIds != null) { - for (Integer cachedRDDId: cachedRDDIds) { - jc.sc().sc().unpersistRDD(cachedRDDId, false); - } - } - } - private void monitorJob(JavaFutureAction job, - SparkCounters sparkCounters, Set cachedRDDIds) { + SparkCounters sparkCounters) { jobs.add(job); if (!jc.getMonitoredJobs().containsKey(req.id)) { jc.getMonitoredJobs().put(req.id, new CopyOnWriteArrayList>()); } jc.getMonitoredJobs().get(req.id).add(job); this.sparkCounters = sparkCounters; - this.cachedRDDIds = cachedRDDIds; protocol.jobSubmitted(req.id, job.jobIds().get(0)); } diff --git spark-client/src/test/java/org/apache/hive/spark/client/TestSparkClient.java spark-client/src/test/java/org/apache/hive/spark/client/TestSparkClient.java index d33ad7e..df91079 100644 --- spark-client/src/test/java/org/apache/hive/spark/client/TestSparkClient.java +++ spark-client/src/test/java/org/apache/hive/spark/client/TestSparkClient.java @@ -319,7 +319,7 @@ public Integer call(JobContext jc) throws Exception { public void call(Integer l) throws Exception { } - }), null, null); + }), null); future.get(TIMEOUT, TimeUnit.SECONDS); @@ -380,7 +380,7 @@ public String call(JobContext jc) { counters.createCounter("group2", "counter2"); jc.monitor(jc.sc().parallelize(Arrays.asList(1, 2, 3, 4, 5), 5).foreachAsync(this), - counters, null); + counters); return null; }