/** * 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 * <p/> * http://www.apache.org/licenses/LICENSE-2.0 * <p/> * 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.hive.spark.counter; import java.io.Serializable; import java.util.HashMap; import java.util.Map; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import org.apache.spark.api.java.JavaSparkContext; /** * SparkCounters is used to collect Hive operator metric through Spark accumulator. There are few * limitation of Spark accumulator, like: * 1. accumulator should be created at Spark context side. * 2. Spark tasks can only increment metric count. * 3. User can only get accumulator value at Spark context side. * These Spark Counter API is designed to fit into Hive requirement, while with several access * restriction due to Spark accumulator previous mentioned: * 1. Counter should be created on driver side if it would be accessed in task. * 2. increment could only be invoked task side. * 3. Hive could only get Counter value at driver side. */ public class SparkCounters implements Serializable { private static final long serialVersionUID = 1L; private static final Logger LOG = LoggerFactory.getLogger(SparkCounters.class); private Map<String, SparkCounterGroup> sparkCounterGroups; private final transient JavaSparkContext javaSparkContext; private SparkCounters() { this(null); } public SparkCounters(JavaSparkContext javaSparkContext) { this.javaSparkContext = javaSparkContext; this.sparkCounterGroups = new HashMap<String, SparkCounterGroup>(); } public void createCounter(Enum<?> key) { createCounter(key.getDeclaringClass().getName(), key.name()); } public void createCounter(String groupName, Enum<?> key) { createCounter(groupName, key.name(), 0L); } public void createCounter(String groupName, String counterName) { createCounter(groupName, counterName, 0L); } public void createCounter(String groupName, String counterName, long initValue) { getGroup(groupName).createCounter(counterName, initValue); } public void increment(Enum<?> key, long incrValue) { increment(key.getDeclaringClass().getName(), key.name(), incrValue); } public void increment(String groupName, String counterName, long value) { SparkCounter counter = getGroup(groupName).getCounter(counterName); if (counter == null) { LOG.error( String.format("counter[%s, %s] has not initialized before.", groupName, counterName)); } else { counter.increment(value); } } public long getValue(String groupName, String counterName) { SparkCounter counter = getGroup(groupName).getCounter(counterName); if (counter == null) { LOG.error( String.format("counter[%s, %s] has not initialized before.", groupName, counterName)); return 0; } else { return counter.getValue(); } } public SparkCounter getCounter(String groupName, String counterName) { return getGroup(groupName).getCounter(counterName); } public SparkCounter getCounter(Enum<?> key) { return getCounter(key.getDeclaringClass().getName(), key.name()); } private SparkCounterGroup getGroup(String groupName) { SparkCounterGroup group = sparkCounterGroups.get(groupName); if (group == null) { group = new SparkCounterGroup(groupName, groupName, javaSparkContext); sparkCounterGroups.put(groupName, group); } return group; } public Map<String, SparkCounterGroup> getSparkCounterGroups() { return sparkCounterGroups; } @Override public String toString() { StringBuilder sb = new StringBuilder(); Map<String, SparkCounterGroup> groups = getSparkCounterGroups(); if (groups != null) { for (Map.Entry<String, SparkCounterGroup> groupEntry : groups.entrySet()) { String groupName = groupEntry.getKey(); SparkCounterGroup group = groupEntry.getValue(); sb.append(groupName).append("\n"); Map<String, SparkCounter> counters = group.getSparkCounters(); for (Map.Entry<String, SparkCounter> counterEntry : counters.entrySet()) { String counterName = counterEntry.getKey(); SparkCounter counter = counterEntry.getValue(); sb.append("\t") .append(counterName) .append(": ") .append(counter.getValue()) .append("\n"); } } } return sb.toString(); } /** * Create a snapshot of the current counters to send back to the client. This copies the values * of all current counters into a new SparkCounters instance that cannot be used to update the * counters, but will serialize cleanly when sent back to the RSC client. */ public SparkCounters snapshot() { SparkCounters snapshot = new SparkCounters(); for (SparkCounterGroup group : sparkCounterGroups.values()) { SparkCounterGroup groupSnapshot = group.snapshot(); snapshot.sparkCounterGroups.put(groupSnapshot.getGroupName(), groupSnapshot); } return snapshot; } }