Package org.apache.tez.mapreduce.protos.MRRuntimeProtos

Examples of org.apache.tez.mapreduce.protos.MRRuntimeProtos.MRInputUserPayloadProto

697
698
699
700
701
702
703
704
705
706
707
708
709
  // This class should not be used by more than one test in a single run, since
  // the path it writes to is not dynamic.
  private static String RELOCALIZATION_TEST_CLASS_NAME = "AMClassloadTestDummyClass";
  public static class MRInputAMSplitGeneratorRelocalizationTest extends MRInputAMSplitGenerator {
    public List<Event> initialize(TezRootInputInitializerContext rootInputContext)  throws Exception {
      MRInputUserPayloadProto userPayloadProto = MRHelpers
          .parseMRInputPayload(rootInputContext.getUserPayload());
      Configuration conf = MRHelpers.createConfFromByteString(userPayloadProto
          .getConfigurationBytes());

      try {
        RuntimeUtils.getClazz(RELOCALIZATION_TEST_CLASS_NAME);
        LOG.info("Class found");
View Full Code Here
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
      throws Exception {
    Stopwatch sw = null;
    if (LOG.isDebugEnabled()) {
      sw = new Stopwatch().start();
    }
    MRInputUserPayloadProto userPayloadProto = MRHelpers
        .parseMRInputPayload(rootInputContext.getUserPayload());
    if (LOG.isDebugEnabled()) {
      sw.stop();
      LOG.debug("Time to parse MRInput payload into prot: "
          + sw.elapsedMillis());
    }
    if (LOG.isDebugEnabled()) {
      sw.reset().start();
    }
    Configuration conf = MRHelpers.createConfFromByteString(userPayloadProto
        .getConfigurationBytes());
   
    sendSerializedEvents = conf.getBoolean(
        MRJobConfig.MR_TEZ_INPUT_INITIALIZER_SERIALIZE_EVENT_PAYLAOD,
        MRJobConfig.MR_TEZ_INPUT_INITIALIZER_SERIALIZE_EVENT_PAYLAOD_DEFAULT);
    LOG.info("Emitting serialized splits: " + sendSerializedEvents);
    if (LOG.isDebugEnabled()) {
      sw.stop();
      LOG.debug("Time converting ByteString to configuration: " + sw.elapsedMillis());
    }

    if (LOG.isDebugEnabled()) {
      sw.reset().start();
    }

    int totalResource = rootInputContext.getTotalAvailableResource().getMemory();
    int taskResource = rootInputContext.getVertexTaskResource().getMemory();
    float waves = conf.getFloat(
        TezConfiguration.TEZ_AM_GROUPING_SPLIT_WAVES,
        TezConfiguration.TEZ_AM_GROUPING_SPLIT_WAVES_DEFAULT);

    int numTasks = (int)((totalResource*waves)/taskResource);

    LOG.info("Input " + rootInputContext.getInputName() + " asking for " + numTasks
        + " tasks. Headroom: " + totalResource + " Task Resource: "
        + taskResource + " waves: " + waves);

    // Read all credentials into the credentials instance stored in JobConf.
    JobConf jobConf = new JobConf(conf);
    jobConf.getCredentials().mergeAll(UserGroupInformation.getCurrentUser().getCredentials());

    InputSplitInfoMem inputSplitInfo = null;
    String realInputFormatName = userPayloadProto.getInputFormatName();
    if ( realInputFormatName != null && !realInputFormatName.isEmpty()) {
      // split grouping on the AM
      if (jobConf.getUseNewMapper()) {
        LOG.info("Grouping mapreduce api input splits");
        Job job = Job.getInstance(jobConf);
View Full Code Here
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
  @Override
  public List<Event> initialize(TezInputContext inputContext) throws IOException {
    this.inputContext = inputContext;
    this.inputContext.requestInitialMemory(0l, null); //mandatory call
    this.inputContext.inputIsReady();
    MRInputUserPayloadProto mrUserPayload =
      MRHelpers.parseMRInputPayload(inputContext.getUserPayload());
    Preconditions.checkArgument(mrUserPayload.hasSplits() == false,
        "Split information not expected in MRInput");
    Configuration conf =
      MRHelpers.createConfFromByteString(mrUserPayload.getConfigurationBytes());
    this.jobConf = new JobConf(conf);
    // Add tokens to the jobConf - in case they are accessed within the RR / IF
    jobConf.getCredentials().mergeAll(UserGroupInformation.getCurrentUser().getCredentials());

    TaskAttemptID taskAttemptId = new TaskAttemptID(
View Full Code Here
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
    rootVertexInitialized = true;
    try {
      // This is using the payload from the RootVertexInitializer corresponding
      // to InputName. Ideally it should be using it's own configuration class - but that
      // means serializing another instance.
      MRInputUserPayloadProto protoPayload =
          MRHelpers.parseMRInputPayload(inputDescriptor.getUserPayload());
      this.conf = MRHelpers.createConfFromByteString(protoPayload.getConfigurationBytes());

      /*
       * Currently in tez, the flow of events is thus: "Generate Splits -> Initialize Vertex"
       * (with parallelism info obtained from the generate splits phase). The generate splits
       * phase groups splits using the TezGroupedSplitsInputFormat. However, for bucket map joins
       * the grouping done by this input format results in incorrect results as the grouper has no
       * knowledge of buckets. So, we initially set the input format to be HiveInputFormat
       * (in DagUtils) for the case of bucket map joins so as to obtain un-grouped splits.
       * We then group the splits corresponding to buckets using the tez grouper which returns
       * TezGroupedSplits.
       */

      if (conf.getBoolean(GROUP_SPLITS, true)) {
        // Changing the InputFormat - so that the correct one is initialized in MRInput.
        this.conf.set("mapred.input.format.class", TezGroupedSplitsInputFormat.class.getName());
        MRInputUserPayloadProto updatedPayload = MRInputUserPayloadProto
            .newBuilder(protoPayload)
            .setConfigurationBytes(MRHelpers.createByteStringFromConf(conf))
            .build();
        inputDescriptor.setUserPayload(updatedPayload.toByteArray());
      }
    } catch (IOException e) {
      e.printStackTrace();
      throw new RuntimeException(e);
    }
View Full Code Here
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
  public List<Event> initialize() throws Exception {
    InputInitializerContext rootInputContext = getContext();

    context = rootInputContext;

    MRInputUserPayloadProto userPayloadProto =
        MRInputHelpers.parseMRInputPayload(rootInputContext.getInputUserPayload());

    Configuration conf =
        TezUtils.createConfFromByteString(userPayloadProto.getConfigurationBytes());

    boolean sendSerializedEvents =
        conf.getBoolean("mapreduce.tez.input.initializer.serialize.event.payload", true);

    // Read all credentials into the credentials instance stored in JobConf.
    JobConf jobConf = new JobConf(conf);
    ShimLoader.getHadoopShims().getMergedCredentials(jobConf);

    MapWork work = Utilities.getMapWork(jobConf);

    // perform dynamic partition pruning
    pruner.prune(work, jobConf, context);

    InputSplitInfoMem inputSplitInfo = null;
    String realInputFormatName = conf.get("mapred.input.format.class");
    boolean groupingEnabled = userPayloadProto.getGroupingEnabled();
    if (groupingEnabled) {
      // Need to instantiate the realInputFormat
      InputFormat<?, ?> inputFormat =
          (InputFormat<?, ?>) ReflectionUtils.newInstance(Class.forName(realInputFormatName),
              jobConf);
View Full Code Here
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
    try {
      // This is using the payload from the RootVertexInitializer corresponding
      // to InputName. Ideally it should be using it's own configuration class -
      // but that
      // means serializing another instance.
      MRInputUserPayloadProto protoPayload =
          MRInputHelpers.parseMRInputPayload(inputDescriptor.getUserPayload());
      this.conf = TezUtils.createConfFromByteString(protoPayload.getConfigurationBytes());

      /*
       * Currently in tez, the flow of events is thus:
       * "Generate Splits -> Initialize Vertex" (with parallelism info obtained
       * from the generate splits phase). The generate splits phase groups
       * splits using the TezGroupedSplitsInputFormat. However, for bucket map
       * joins the grouping done by this input format results in incorrect
       * results as the grouper has no knowledge of buckets. So, we initially
       * set the input format to be HiveInputFormat (in DagUtils) for the case
       * of bucket map joins so as to obtain un-grouped splits. We then group
       * the splits corresponding to buckets using the tez grouper which returns
       * TezGroupedSplits.
       */

      // This assumes that Grouping will always be used.
      // Enabling grouping on the payload.
      MRInputUserPayloadProto updatedPayload =
          MRInputUserPayloadProto.newBuilder(protoPayload).setGroupingEnabled(true).build();
      inputDescriptor.setUserPayload(UserPayload.create(updatedPayload.toByteString().asReadOnlyByteBuffer()));
    } catch (IOException e) {
      e.printStackTrace();
      throw new RuntimeException(e);
    }

View Full Code Here
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
    rootVertexInitialized = true;
    try {
      // This is using the payload from the RootVertexInitializer corresponding
      // to InputName. Ideally it should be using it's own configuration class - but that
      // means serializing another instance.
      MRInputUserPayloadProto protoPayload =
          MRHelpers.parseMRInputPayload(inputDescriptor.getUserPayload());
      this.conf = MRHelpers.createConfFromByteString(protoPayload.getConfigurationBytes());

      /*
       * Currently in tez, the flow of events is thus: "Generate Splits -> Initialize Vertex"
       * (with parallelism info obtained from the generate splits phase). The generate splits
       * phase groups splits using the TezGroupedSplitsInputFormat. However, for bucket map joins
       * the grouping done by this input format results in incorrect results as the grouper has no
       * knowledge of buckets. So, we initially set the input format to be HiveInputFormat
       * (in DagUtils) for the case of bucket map joins so as to obtain un-grouped splits.
       * We then group the splits corresponding to buckets using the tez grouper which returns
       * TezGroupedSplits.
       */

      if (conf.getBoolean(GROUP_SPLITS, true)) {
        // Changing the InputFormat - so that the correct one is initialized in MRInput.
        this.conf.set("mapred.input.format.class", TezGroupedSplitsInputFormat.class.getName());
        MRInputUserPayloadProto updatedPayload = MRInputUserPayloadProto
            .newBuilder(protoPayload)
            .setConfigurationBytes(MRHelpers.createByteStringFromConf(conf))
            .build();
        inputDescriptor.setUserPayload(updatedPayload.toByteArray());
      }
    } catch (IOException e) {
      e.printStackTrace();
      throw new RuntimeException(e);
    }
View Full Code Here
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
    rootVertexInitialized = true;
    try {
      // This is using the payload from the RootVertexInitializer corresponding
      // to InputName. Ideally it should be using it's own configuration class - but that
      // means serializing another instance.
      MRInputUserPayloadProto protoPayload =
          MRHelpers.parseMRInputPayload(inputDescriptor.getUserPayload());
      this.conf = MRHelpers.createConfFromByteString(protoPayload.getConfigurationBytes());

      /*
       * Currently in tez, the flow of events is thus: "Generate Splits -> Initialize Vertex"
       * (with parallelism info obtained from the generate splits phase). The generate splits
       * phase groups splits using the TezGroupedSplitsInputFormat. However, for bucket map joins
       * the grouping done by this input format results in incorrect results as the grouper has no
       * knowledge of buckets. So, we initially set the input format to be HiveInputFormat
       * (in DagUtils) for the case of bucket map joins so as to obtain un-grouped splits.
       * We then group the splits corresponding to buckets using the tez grouper which returns
       * TezGroupedSplits.
       */

      // This assumes that Grouping will always be used.
      // Changing the InputFormat - so that the correct one is initialized in MRInput.
      this.conf.set("mapred.input.format.class", TezGroupedSplitsInputFormat.class.getName());
      MRInputUserPayloadProto updatedPayload = MRInputUserPayloadProto
          .newBuilder(protoPayload)
          .setConfigurationBytes(MRHelpers.createByteStringFromConf(conf))
          .build();
      inputDescriptor.setUserPayload(updatedPayload.toByteArray());
    } catch (IOException e) {
      e.printStackTrace();
      throw new RuntimeException(e);
    }
    boolean dataInformationEventSeen = false;
View Full Code Here
669
670
671
672
673
674
675
676
677
678
679
680
681
  // This class should not be used by more than one test in a single run, since
  // the path it writes to is not dynamic.
  private static String RELOCALIZATION_TEST_CLASS_NAME = "AMClassloadTestDummyClass";
  public static class MRInputAMSplitGeneratorRelocalizationTest extends MRInputAMSplitGenerator {
    public List<Event> initialize(TezRootInputInitializerContext rootInputContext)  throws Exception {
      MRInputUserPayloadProto userPayloadProto = MRHelpers
          .parseMRInputPayload(rootInputContext.getUserPayload());
      Configuration conf = MRHelpers.createConfFromByteString(userPayloadProto
          .getConfigurationBytes());

      try {
        RuntimeUtils.getClazz(RELOCALIZATION_TEST_CLASS_NAME);
        LOG.info("Class found");
View Full Code Here
TOP

Related Classes of org.apache.tez.mapreduce.protos.MRRuntimeProtos.MRInputUserPayloadProto

  • org.apache.hadoop.hive.ql.exec.tez.CustomPartitionVertex
  • org.apache.hadoop.hive.ql.exec.tez.HiveSplitGenerator
  • org.apache.tez.mapreduce.common.MRInputAMSplitGenerator
  • org.apache.tez.mapreduce.common.MRInputSplitDistributor
  • org.apache.tez.mapreduce.input.MRInput
  • org.apache.tez.mapreduce.TestMRRJobsDAGApi$MRInputAMSplitGeneratorRelocalizationTest
    • Copyright © 2018 www.massapicom. All rights reserved.
    All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.