[GitHub] [druid] jon-wei commented on a change in pull request #10419: Automatically determine numShards for parallel ingestion hash partitioning

[GitBox <gi...@apache.org>]

jon-wei commented on a change in pull request #10419:
URL: https://github.com/apache/druid/pull/10419#discussion_r493117965



##########
File path: indexing-service/src/main/java/org/apache/druid/indexing/common/task/batch/parallel/PartialDimensionCardinalityTask.java
##########
@@ -0,0 +1,258 @@
+/*
+ * 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.druid.indexing.common.task.batch.parallel;
+
+import com.fasterxml.jackson.annotation.JacksonInject;
+import com.fasterxml.jackson.annotation.JsonCreator;
+import com.fasterxml.jackson.annotation.JsonProperty;
+import com.fasterxml.jackson.core.JsonProcessingException;
+import com.fasterxml.jackson.databind.ObjectMapper;
+import com.google.common.base.Preconditions;
+import org.apache.druid.data.input.HandlingInputRowIterator;
+import org.apache.druid.data.input.InputFormat;
+import org.apache.druid.data.input.InputRow;
+import org.apache.druid.data.input.InputSource;
+import org.apache.druid.hll.HyperLogLogCollector;
+import org.apache.druid.indexer.TaskStatus;
+import org.apache.druid.indexer.partitions.HashedPartitionsSpec;
+import org.apache.druid.indexing.common.TaskToolbox;
+import org.apache.druid.indexing.common.actions.TaskActionClient;
+import org.apache.druid.indexing.common.task.AbstractBatchIndexTask;
+import org.apache.druid.indexing.common.task.ClientBasedTaskInfoProvider;
+import org.apache.druid.indexing.common.task.IndexTask;
+import org.apache.druid.indexing.common.task.TaskResource;
+import org.apache.druid.indexing.common.task.batch.parallel.iterator.DefaultIndexTaskInputRowIteratorBuilder;
+import org.apache.druid.java.util.common.logger.Logger;
+import org.apache.druid.java.util.common.parsers.CloseableIterator;
+import org.apache.druid.segment.incremental.ParseExceptionHandler;
+import org.apache.druid.segment.incremental.RowIngestionMeters;
+import org.apache.druid.segment.indexing.DataSchema;
+import org.apache.druid.segment.indexing.granularity.GranularitySpec;
+import org.apache.druid.timeline.partition.HashBasedNumberedShardSpec;
+import org.joda.time.DateTime;
+import org.joda.time.Interval;
+
+import javax.annotation.Nullable;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+public class PartialDimensionCardinalityTask extends PerfectRollupWorkerTask
+{
+  public static final String TYPE = "partial_dimension_cardinality";
+  private static final Logger LOG = new Logger(PartialDimensionCardinalityTask.class);
+
+  private final int numAttempts;
+  private final ParallelIndexIngestionSpec ingestionSchema;
+  private final String supervisorTaskId;
+
+  private final ObjectMapper jsonMapper;
+
+  @JsonCreator
+  PartialDimensionCardinalityTask(
+      // id shouldn't be null except when this task is created by ParallelIndexSupervisorTask
+      @JsonProperty("id") @Nullable String id,
+      @JsonProperty("groupId") final String groupId,
+      @JsonProperty("resource") final TaskResource taskResource,
+      @JsonProperty("supervisorTaskId") final String supervisorTaskId,
+      @JsonProperty("numAttempts") final int numAttempts, // zero-based counting
+      @JsonProperty("spec") final ParallelIndexIngestionSpec ingestionSchema,
+      @JsonProperty("context") final Map<String, Object> context,
+      @JacksonInject ObjectMapper jsonMapper
+  )
+  {
+    super(
+        getOrMakeId(id, TYPE, ingestionSchema.getDataSchema().getDataSource()),
+        groupId,
+        taskResource,
+        ingestionSchema.getDataSchema(),
+        ingestionSchema.getTuningConfig(),
+        context
+    );
+
+    Preconditions.checkArgument(
+        ingestionSchema.getTuningConfig().getPartitionsSpec() instanceof HashedPartitionsSpec,
+        "%s partitionsSpec required",
+        HashedPartitionsSpec.NAME
+    );
+
+    this.numAttempts = numAttempts;
+    this.ingestionSchema = ingestionSchema;
+    this.supervisorTaskId = supervisorTaskId;
+    this.jsonMapper = jsonMapper;
+  }
+
+  @JsonProperty
+  private int getNumAttempts()
+  {
+    return numAttempts;
+  }
+
+  @JsonProperty("spec")
+  private ParallelIndexIngestionSpec getIngestionSchema()
+  {
+    return ingestionSchema;
+  }
+
+  @JsonProperty
+  private String getSupervisorTaskId()
+  {
+    return supervisorTaskId;
+  }
+
+  @Override
+  public String getType()
+  {
+    return TYPE;
+  }
+
+  @Override
+  public boolean isReady(TaskActionClient taskActionClient) throws Exception
+  {
+    return tryTimeChunkLock(
+        taskActionClient,
+        getIngestionSchema().getDataSchema().getGranularitySpec().inputIntervals()
+    );
+  }
+
+  @Override
+  public TaskStatus runTask(TaskToolbox toolbox) throws Exception
+  {
+    DataSchema dataSchema = ingestionSchema.getDataSchema();
+    GranularitySpec granularitySpec = dataSchema.getGranularitySpec();
+    ParallelIndexTuningConfig tuningConfig = ingestionSchema.getTuningConfig();
+
+    HashedPartitionsSpec partitionsSpec = (HashedPartitionsSpec) tuningConfig.getPartitionsSpec();
+    Preconditions.checkNotNull(partitionsSpec, "partitionsSpec required in tuningConfig");
+
+    List<String> partitionDimensions = partitionsSpec.getPartitionDimensions();
+    if (partitionDimensions == null) {
+      partitionDimensions = HashBasedNumberedShardSpec.DEFAULT_PARTITION_DIMENSIONS;
+    }
+
+    InputSource inputSource = ingestionSchema.getIOConfig().getNonNullInputSource(
+        ingestionSchema.getDataSchema().getParser()
+    );
+    InputFormat inputFormat = inputSource.needsFormat()
+                              ? ParallelIndexSupervisorTask.getInputFormat(ingestionSchema)
+                              : null;
+    final RowIngestionMeters buildSegmentsMeters = toolbox.getRowIngestionMetersFactory().createRowIngestionMeters();
+    final ParseExceptionHandler parseExceptionHandler = new ParseExceptionHandler(
+        buildSegmentsMeters,
+        tuningConfig.isLogParseExceptions(),
+        tuningConfig.getMaxParseExceptions(),
+        tuningConfig.getMaxSavedParseExceptions()
+    );
+
+    try (
+        final CloseableIterator<InputRow> inputRowIterator = AbstractBatchIndexTask.inputSourceReader(
+            toolbox.getIndexingTmpDir(),
+            dataSchema,
+            inputSource,
+            inputFormat,
+            AbstractBatchIndexTask.defaultRowFilter(granularitySpec),
+            buildSegmentsMeters,
+            parseExceptionHandler
+        );
+        HandlingInputRowIterator iterator =
+            new DefaultIndexTaskInputRowIteratorBuilder()
+                .delegate(inputRowIterator)
+                .granularitySpec(granularitySpec)
+                .build()
+    ) {
+      Map<Interval, byte[]> cardinalities = determineCardinalities(
+          iterator,
+          granularitySpec,
+          partitionDimensions
+      );
+
+      sendReport(
+          toolbox,
+          new DimensionCardinalityReport(getId(), cardinalities)
+      );
+    }
+
+    return TaskStatus.success(getId());
+  }
+
+  private Map<Interval, byte[]> determineCardinalities(
+      HandlingInputRowIterator inputRowIterator,
+      GranularitySpec granularitySpec,
+      List<String> partitionDimensions
+  )
+  {
+    Map<Interval, HyperLogLogCollector> intervalToCardinalities = new HashMap<>();
+    while (inputRowIterator.hasNext()) {
+      InputRow inputRow = inputRowIterator.next();
+      if (inputRow == null) {
+        continue;
+      }
+
+      DateTime timestamp = inputRow.getTimestamp();
+
+      //noinspection OptionalGetWithoutIsPresent (InputRowIterator returns rows with present intervals)
+      Interval interval = granularitySpec.bucketInterval(timestamp).get();
+
+      LOG.info("TS: " + timestamp + " INTV: " + interval + " GSC: " + granularitySpec.getClass());
+
+      HyperLogLogCollector hllCollector = intervalToCardinalities.computeIfAbsent(

Review comment:
       > Using HllSketch would mean that the implementation for parallel ingestion is different from the one for sequential ingestion though.
   
   That was my reasoning for using `HyperLogLogCollector`, but I think it makes sense to change these to use `HllSketch`. I'll update this one to use `HllSketch` and a follow-on could be to do the same for `IndexTask`.
   
   




----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org