HDFS-11722. Change Datanode file IO profiling sampling to percentage. Contributed by Hanisha Koneru.

hadoop-common-project/hadoop-common/src/site/markdown/Metrics.md

@@ -334,8 +334,8 @@ FsVolume
 
 Per-volume metrics contain Datanode Volume IO related statistics. Per-volume
 metrics are off by default. They can be enabled by setting `dfs.datanode
-.fileio.profiling.sampling.fraction` to a fraction between 0.0 and 1.0.
-Setting this value to 0.0 would mean profiling is not enabled. But enabling
+.fileio.profiling.percentage.fraction` to an integer value between 1 and 100.
+Setting this value to 0 would mean profiling is not enabled. But enabling
 per-volume metrics may have a performance impact. Each metrics record
 contains tags such as Hostname as additional information along with metrics.
 

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/DFSConfigKeys.java

@@ -731,10 +731,10 @@ public class DFSConfigKeys extends CommonConfigurationKeys {
   public static final boolean
       DFS_DATANODE_ENABLE_FILEIO_FAULT_INJECTION_DEFAULT = false;
   public static final String
-      DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_KEY =
-      "dfs.datanode.fileio.profiling.sampling.fraction";
-  public static final double
-      DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_DEFAULT = 0.0;
+      DFS_DATANODE_FILEIO_PROFILING_SAMPLING_PERCENTAGE_KEY =
+      "dfs.datanode.fileio.profiling.sampling.percentage";
+  public static final int
+      DFS_DATANODE_FILEIO_PROFILING_SAMPLING_PERCENTAGE_DEFAULT = 0;
 
   //Keys with no defaults
   public static final String  DFS_DATANODE_PLUGINS_KEY = "dfs.datanode.plugins";

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/DatanodeManager.java

@@ -215,9 +215,10 @@ public class DatanodeManager {
     this.dataNodePeerStatsEnabled = conf.getBoolean(
         DFSConfigKeys.DFS_DATANODE_PEER_STATS_ENABLED_KEY,
         DFSConfigKeys.DFS_DATANODE_PEER_STATS_ENABLED_DEFAULT);
-    this.dataNodeDiskStatsEnabled = Util.isDiskStatsEnabled(conf.getDouble(
-        DFSConfigKeys.DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_KEY,
-        DFSConfigKeys.DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_DEFAULT));
+    this.dataNodeDiskStatsEnabled = Util.isDiskStatsEnabled(conf.getInt(
+        DFSConfigKeys.DFS_DATANODE_FILEIO_PROFILING_SAMPLING_PERCENTAGE_KEY,
+        DFSConfigKeys.
+            DFS_DATANODE_FILEIO_PROFILING_SAMPLING_PERCENTAGE_DEFAULT));
 
     final Timer timer = new Timer();
     this.slowPeerTracker = dataNodePeerStatsEnabled ?

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/common/Util.java

@@ -389,17 +389,17 @@ public final class Util {
     return addrsList;
   }
 
-  public static boolean isDiskStatsEnabled(double fileIOSamplingFraction) {
+  public static boolean isDiskStatsEnabled(int fileIOSamplingPercentage) {
     final boolean isEnabled;
-    if (fileIOSamplingFraction < 0.000001) {
+    if (fileIOSamplingPercentage <= 0) {
       LOG.info(DFSConfigKeys
-          .DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_KEY + " set to "
-          + fileIOSamplingFraction + ". Disabling file IO profiling");
+          .DFS_DATANODE_FILEIO_PROFILING_SAMPLING_PERCENTAGE_KEY + " set to "
+          + fileIOSamplingPercentage + ". Disabling file IO profiling");
       isEnabled = false;
     } else {
       LOG.info(DFSConfigKeys
-          .DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_KEY + " set to "
-          + fileIOSamplingFraction + ". Enabling file IO profiling");
+          .DFS_DATANODE_FILEIO_PROFILING_SAMPLING_PERCENTAGE_KEY + " set to "
+          + fileIOSamplingPercentage + ". Enabling file IO profiling");
       isEnabled = true;
     }
 

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/DNConf.java

@@ -175,9 +175,10 @@ public class DNConf {
     this.peerStatsEnabled = getConf().getBoolean(
         DFSConfigKeys.DFS_DATANODE_PEER_STATS_ENABLED_KEY,
         DFSConfigKeys.DFS_DATANODE_PEER_STATS_ENABLED_DEFAULT);
-    this.diskStatsEnabled = Util.isDiskStatsEnabled(getConf().getDouble(
-        DFSConfigKeys.DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_KEY,
-        DFSConfigKeys.DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_DEFAULT));
+    this.diskStatsEnabled = Util.isDiskStatsEnabled(getConf().getInt(
+        DFSConfigKeys.DFS_DATANODE_FILEIO_PROFILING_SAMPLING_PERCENTAGE_KEY,
+        DFSConfigKeys.
+            DFS_DATANODE_FILEIO_PROFILING_SAMPLING_PERCENTAGE_DEFAULT));
     this.outliersReportIntervalMs = getConf().getTimeDuration(
         DFS_DATANODE_OUTLIERS_REPORT_INTERVAL_KEY,
         DFS_DATANODE_OUTLIERS_REPORT_INTERVAL_DEFAULT,

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/FileIoProvider.java

@@ -62,8 +62,8 @@ import static org.apache.hadoop.hdfs.server.datanode.FileIoProvider.OPERATION.*;
  *
  * Behavior can be injected into these events by enabling the
  * profiling and/or fault injection event hooks through
- * {@link DFSConfigKeys#DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_KEY} and
- * {@link DFSConfigKeys#DFS_DATANODE_ENABLE_FILEIO_FAULT_INJECTION_KEY}.
+ * {@link DFSConfigKeys#DFS_DATANODE_FILEIO_PROFILING_SAMPLING_PERCENTAGE_KEY}
+ * and {@link DFSConfigKeys#DFS_DATANODE_ENABLE_FILEIO_FAULT_INJECTION_KEY}.
  * These event hooks are disabled by default.
  *
  * Most functions accept an optional {@link FsVolumeSpi} parameter for

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/ProfilingFileIoEvents.java

@@ -44,18 +44,19 @@ class ProfilingFileIoEvents {
 
   public ProfilingFileIoEvents(@Nullable Configuration conf) {
     if (conf != null) {
-      double fileIOSamplingFraction = conf.getDouble(
-          DFSConfigKeys.DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_KEY,
+      int fileIOSamplingPercentage = conf.getInt(
+          DFSConfigKeys.DFS_DATANODE_FILEIO_PROFILING_SAMPLING_PERCENTAGE_KEY,
           DFSConfigKeys
-              .DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_DEFAULT);
-      isEnabled = Util.isDiskStatsEnabled(fileIOSamplingFraction);
-      if (fileIOSamplingFraction > 1) {
+              .DFS_DATANODE_FILEIO_PROFILING_SAMPLING_PERCENTAGE_DEFAULT);
+      isEnabled = Util.isDiskStatsEnabled(fileIOSamplingPercentage);
+      if (fileIOSamplingPercentage > 100) {
         LOG.warn(DFSConfigKeys
-            .DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_KEY +
-            " value cannot be more than 1. Setting value to 1");
-        fileIOSamplingFraction = 1;
+            .DFS_DATANODE_FILEIO_PROFILING_SAMPLING_PERCENTAGE_KEY +
+            " value cannot be more than 100. Setting value to 100");
+        fileIOSamplingPercentage = 100;
       }
-      sampleRangeMax = (int) (fileIOSamplingFraction * Integer.MAX_VALUE);
+      sampleRangeMax = (int) ((double) fileIOSamplingPercentage / 100 *
+          Integer.MAX_VALUE);
     } else {
       isEnabled = false;
       sampleRangeMax = 0;

hadoop-hdfs-project/hadoop-hdfs/src/main/resources/hdfs-default.xml

@@ -2021,6 +2021,17 @@
   </description>
 </property>
 
+<property>
+  <name>dfs.datanode.fileio.profiling.sampling.percentage</name>
+  <value>0</value>
+  <description>
+    This setting controls the percentage of file I/O events which will be
+    profiled for DataNode disk statistics. The default value of 0 disables
+    disk statistics. Set to an integer value between 1 and 100 to enable disk
+    statistics.
+  </description>
+</property>
+
 <property>
   <name>hadoop.user.group.metrics.percentiles.intervals</name>
   <value></value>

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/blockmanagement/TestSlowDiskTracker.java

@@ -28,7 +28,7 @@ import com.fasterxml.jackson.databind.ObjectMapper;
 import com.google.common.collect.ImmutableMap;
 import org.apache.hadoop.conf.Configuration;
 import static org.apache.hadoop.hdfs.DFSConfigKeys
-    .DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_KEY;
+    .DFS_DATANODE_FILEIO_PROFILING_SAMPLING_PERCENTAGE_KEY;
 import static org.apache.hadoop.hdfs.DFSConfigKeys
     .DFS_DATANODE_OUTLIERS_REPORT_INTERVAL_KEY;
 import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_HEARTBEAT_INTERVAL_KEY;
@@ -80,7 +80,7 @@ public class TestSlowDiskTracker {
   static {
     conf = new HdfsConfiguration();
     conf.setLong(DFS_HEARTBEAT_INTERVAL_KEY, 1L);
-    conf.setDouble(DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_KEY, 1.0);
+    conf.setInt(DFS_DATANODE_FILEIO_PROFILING_SAMPLING_PERCENTAGE_KEY, 100);
     conf.setTimeDuration(DFS_DATANODE_OUTLIERS_REPORT_INTERVAL_KEY,
         OUTLIERS_REPORT_INTERVAL, TimeUnit.MILLISECONDS);
   }

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/datanode/TestDataNodeMXBean.java

@@ -218,8 +218,8 @@ public class TestDataNodeMXBean {
   @Test
   public void testDataNodeMXBeanSlowDisksEnabled() throws Exception {
     Configuration conf = new Configuration();
-    conf.setDouble(DFSConfigKeys
-        .DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_KEY, 1.0);
+    conf.setInt(DFSConfigKeys
+        .DFS_DATANODE_FILEIO_PROFILING_SAMPLING_PERCENTAGE_KEY, 100);
     MiniDFSCluster cluster = new MiniDFSCluster.Builder(conf).build();
 
     try {

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/datanode/TestDataNodeVolumeMetrics.java

@@ -121,8 +121,8 @@ public class TestDataNodeVolumeMetrics {
 
   private MiniDFSCluster setupClusterForVolumeMetrics() throws IOException {
     Configuration conf = new HdfsConfiguration();
-    conf.setDouble(DFSConfigKeys
-        .DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_KEY, 1.0);
+    conf.setInt(DFSConfigKeys
+        .DFS_DATANODE_FILEIO_PROFILING_SAMPLING_PERCENTAGE_KEY, 100);
     SimulatedFSDataset.setFactory(conf);
     return new MiniDFSCluster.Builder(conf)
         .numDataNodes(NUM_DATANODES)

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestNameNodeStatusMXBean.java

@@ -105,8 +105,9 @@ public class TestNameNodeStatusMXBean {
   @Test
   public void testNameNodeMXBeanSlowDisksEnabled() throws Exception {
     Configuration conf = new Configuration();
-    conf.setDouble(
-        DFSConfigKeys.DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_KEY, 1.0);
+    conf.setInt(
+        DFSConfigKeys.DFS_DATANODE_FILEIO_PROFILING_SAMPLING_PERCENTAGE_KEY,
+        100);
     conf.setTimeDuration(
         DFSConfigKeys.DFS_DATANODE_OUTLIERS_REPORT_INTERVAL_KEY,
         1000, TimeUnit.MILLISECONDS);

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/tools/TestHdfsConfigFields.java

@@ -105,8 +105,6 @@ public class TestHdfsConfigFields extends TestConfigurationFieldsBase {
         .add(DFSConfigKeys.DFS_NAMENODE_STARTUP_KEY);
     configurationPropsToSkipCompare.add(DFSConfigKeys
         .DFS_DATANODE_ENABLE_FILEIO_FAULT_INJECTION_KEY);
-    configurationPropsToSkipCompare.add(DFSConfigKeys
-        .DFS_DATANODE_FILEIO_PROFILING_SAMPLING_FRACTION_KEY);
 
     // Allocate
     xmlPropsToSkipCompare = new HashSet<String>();