Skip to content

[SPARK-18928] Check TaskContext.isInterrupted() in FileScanRDD, JDBCRDD & UnsafeSorter #16340

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 3 commits into from
Closed

[SPARK-18928] Check TaskContext.isInterrupted() in FileScanRDD, JDBCRDD & UnsafeSorter #16340

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
08f9aa5
Fix FileScanRDD interruption.
JoshRosen Dec 19, 2016
2d43d5a
Fix for JDBCRDD interruption.
JoshRosen Dec 19, 2016
236efe5
Make UnsafeSorterIterator interruptible.
JoshRosen Dec 19, 2016
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 11 additions & 0 deletions core/src/main/java/org/apache/spark/util/collection/unsafe/sort/UnsafeInMemorySorter.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,8 @@

import org.apache.avro.reflect.Nullable;

import org.apache.spark.TaskContext;
import org.apache.spark.TaskKilledException;
import org.apache.spark.memory.MemoryConsumer;
import org.apache.spark.memory.TaskMemoryManager;
import org.apache.spark.unsafe.Platform;
Expand Down Expand Up @@ -253,6 +255,7 @@ public final class SortedIterator extends UnsafeSorterIterator implements Clonea
private long keyPrefix;
private int recordLength;
private long currentPageNumber;
private final TaskContext taskContext = TaskContext.get();

private SortedIterator(int numRecords, int offset) {
this.numRecords = numRecords;
Expand Down Expand Up @@ -283,6 +286,14 @@ public boolean hasNext() {

@Override
public void loadNext() {
// Kill the task in case it has been marked as killed. This logic is from
// InterruptibleIterator, but we inline it here instead of wrapping the iterator in order
// to avoid performance overhead. This check is added here in `loadNext()` instead of in
// `hasNext()` because it's technically possible for the caller to be relying on
// `getNumRecords()` instead of `hasNext()` to know when to stop.
if (taskContext != null && taskContext.isInterrupted()) {
throw new TaskKilledException();
}
// This pointer points to a 4-byte record length, followed by the record's bytes
final long recordPointer = array.get(offset + position);
currentPageNumber = TaskMemoryManager.decodePageNumber(recordPointer);
Expand Down
11 changes: 11 additions & 0 deletions core/src/main/java/org/apache/spark/util/collection/unsafe/sort/UnsafeSorterSpillReader.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,8 @@
import com.google.common.io.Closeables;

import org.apache.spark.SparkEnv;
import org.apache.spark.TaskContext;
import org.apache.spark.TaskKilledException;
import org.apache.spark.io.NioBufferedFileInputStream;
import org.apache.spark.serializer.SerializerManager;
import org.apache.spark.storage.BlockId;
Expand Down Expand Up @@ -51,6 +53,7 @@ public final class UnsafeSorterSpillReader extends UnsafeSorterIterator implemen
private byte[] arr = new byte[1024 * 1024];
private Object baseObject = arr;
private final long baseOffset = Platform.BYTE_ARRAY_OFFSET;
private final TaskContext taskContext = TaskContext.get();
Copy link
Contributor Author

@JoshRosen JoshRosen Dec 19, 2016

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I didn't want to change the constructor, hence this pattern.


public UnsafeSorterSpillReader(
SerializerManager serializerManager,
Expand Down Expand Up @@ -94,6 +97,14 @@ public boolean hasNext() {

@Override
public void loadNext() throws IOException {
// Kill the task in case it has been marked as killed. This logic is from
// InterruptibleIterator, but we inline it here instead of wrapping the iterator in order
// to avoid performance overhead. This check is added here in `loadNext()` instead of in
// `hasNext()` because it's technically possible for the caller to be relying on
// `getNumRecords()` instead of `hasNext()` to know when to stop.
if (taskContext != null && taskContext.isInterrupted()) {
Copy link
Contributor Author

@JoshRosen JoshRosen Dec 19, 2016

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

TaskContext can be null in case this is used on the driver outside of the context of a specific task.

throw new TaskKilledException();
}
recordLength = din.readInt();
keyPrefix = din.readLong();
if (recordLength > arr.length) {
Expand Down
12 changes: 10 additions & 2 deletions sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileScanRDD.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,7 +21,7 @@ import java.io.IOException

import scala.collection.mutable

import org.apache.spark.{Partition => RDDPartition, TaskContext}
import org.apache.spark.{Partition => RDDPartition, TaskContext, TaskKilledException}
import org.apache.spark.deploy.SparkHadoopUtil
import org.apache.spark.rdd.{InputFileBlockHolder, RDD}
import org.apache.spark.sql.SparkSession
Expand Down Expand Up @@ -99,7 +99,15 @@ class FileScanRDD(
private[this] var currentFile: PartitionedFile = null
private[this] var currentIterator: Iterator[Object] = null

def hasNext: Boolean = (currentIterator != null && currentIterator.hasNext) || nextIterator()
def hasNext: Boolean = {
// Kill the task in case it has been marked as killed. This logic is from
// InterruptibleIterator, but we inline it here instead of wrapping the iterator in order
// to avoid performance overhead.
if (context.isInterrupted()) {
throw new TaskKilledException
}
(currentIterator != null && currentIterator.hasNext) || nextIterator()
}
def next(): Object = {
val nextElement = currentIterator.next()
// TODO: we should have a better separation of row based and batch based scan, so that we
Expand Down
5 changes: 3 additions & 2 deletions sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/jdbc/JDBCRDD.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,7 +23,7 @@ import scala.util.control.NonFatal

import org.apache.commons.lang3.StringUtils

import org.apache.spark.{Partition, SparkContext, TaskContext}
import org.apache.spark.{InterruptibleIterator, Partition, SparkContext, TaskContext}
import org.apache.spark.internal.Logging
import org.apache.spark.rdd.RDD
import org.apache.spark.sql.catalyst.InternalRow
Expand Down Expand Up @@ -301,6 +301,7 @@ private[jdbc] class JDBCRDD(
rs = stmt.executeQuery()
val rowsIterator = JdbcUtils.resultSetToSparkInternalRows(rs, schema, inputMetrics)

CompletionIterator[InternalRow, Iterator[InternalRow]](rowsIterator, close())
CompletionIterator[InternalRow, Iterator[InternalRow]](
new InterruptibleIterator(context, rowsIterator), close())
Copy link
Contributor Author

@JoshRosen JoshRosen Dec 19, 2016

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I suppose I could also have added the check into resultSetToSparkInternalRows but that function is exposed for use outside of Spark internals. Also, I think that JDBCRDD is going to be slow enough that the performance impact here shouldn't be noticeable.

}
}