Merge branch 'apache:master' into HDDS-11243

.github/workflows/ci.yml

@@ -31,6 +31,8 @@ env:
   # Minimum required Java version for running Ozone is defined in pom.xml (javac.version).
   TEST_JAVA_VERSION: 17 # JDK version used by CI build and tests; should match the JDK version in apache/ozone-runner image
   MAVEN_OPTS: -Dhttp.keepAlive=false -Dmaven.wagon.http.pool=false -Dmaven.wagon.http.retryHandler.class=standard -Dmaven.wagon.http.retryHandler.count=3
+  OZONE_RUNNER_IMAGE: ghcr.io/apache/ozone-runner
+  OZONE_RUNNER_VERSION: 20241108-jdk17-1
   OZONE_WITH_COVERAGE: ${{ github.event_name == 'push' }}
 jobs:
   build-info:

hadoop-hdds/common/src/main/java/org/apache/hadoop/hdds/HddsConfigKeys.java

@@ -405,4 +405,7 @@ private HddsConfigKeys() {
       "hdds.datanode.slow.op.warning.threshold";
   public static final String HDDS_DATANODE_SLOW_OP_WARNING_THRESHOLD_DEFAULT =
       "500ms";
+
+  public static final String OZONE_DATANODE_IO_METRICS_PERCENTILES_INTERVALS_SECONDS_KEY =
+      "ozone.volume.io.percentiles.intervals.seconds";
 }

hadoop-hdds/common/src/main/resources/ozone-default.xml

@@ -4553,4 +4553,16 @@
       maximum number of buckets across all volumes.
     </description>
   </property>
+
+  <property>
+    <name>ozone.volume.io.percentiles.intervals.seconds</name>
+    <value>60</value>
+    <tag>OZONE, DATANODE</tag>
+    <description>
+      This setting specifies the interval (in seconds) for monitoring percentile performance metrics.
+      It helps in tracking the read and write performance of DataNodes in real-time,
+      allowing for better identification and analysis of performance issues.
+    </description>
+  </property>
+
 </configuration>

hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/DNMXBean.java

@@ -26,4 +26,32 @@
  */
 @InterfaceAudience.Private
 public interface DNMXBean extends ServiceRuntimeInfo {
+
+  /**
+   * Gets the datanode hostname.
+   *
+   * @return the datanode hostname for the datanode.
+   */
+  String getHostname();
+
+  /**
+   * Gets the client rpc port.
+   *
+   * @return the client rpc port
+   */
+  String getClientRpcPort();
+
+  /**
+   * Gets the http port.
+   *
+   * @return the http port
+   */
+  String getHttpPort();
+
+  /**
+   * Gets the https port.
+   *
+   * @return the http port
+   */
+  String getHttpsPort();
 }

hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/DNMXBeanImpl.java

@@ -25,8 +25,53 @@
  * This is the JMX management class for DN information.
  */
 public class DNMXBeanImpl extends ServiceRuntimeInfoImpl implements DNMXBean {
-  public DNMXBeanImpl(
-      VersionInfo versionInfo) {
+
+  private String hostName;
+  private String clientRpcPort;
+  private String httpPort;
+  private String httpsPort;
+
+  public DNMXBeanImpl(VersionInfo versionInfo) {
     super(versionInfo);
   }
+
+  @Override
+  public String getHostname() {
+    return hostName;
+  }
+
+  @Override
+  public String getClientRpcPort() {
+    return clientRpcPort;
+  }
+
+  @Override
+  public String getHttpPort() {
+    return httpPort;
+  }
+
+  @Override
+  public String getHttpsPort() {
+    return httpsPort;
+  }
+
+  public void setHttpPort(String httpPort) {
+    this.httpPort = httpPort;
+  }
+
+  public void setHostName(String hostName) {
+    this.hostName = hostName;
+  }
+
+  public void setClientRpcPort(String rpcPort) {
+    this.clientRpcPort = rpcPort;
+  }
+
+  public String getHostName() {
+    return hostName;
+  }
+
+  public void setHttpsPort(String httpsPort) {
+    this.httpsPort = httpsPort;
+  }
 }

hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/HddsDatanodeService.java

@@ -228,6 +228,7 @@ public String getNamespace() {
       String ip = InetAddress.getByName(hostname).getHostAddress();
       datanodeDetails = initializeDatanodeDetails();
       datanodeDetails.setHostName(hostname);
+      serviceRuntimeInfo.setHostName(hostname);
       datanodeDetails.setIpAddress(ip);
       datanodeDetails.setVersion(
           HddsVersionInfo.HDDS_VERSION_INFO.getVersion());
@@ -300,23 +301,30 @@ public String getNamespace() {
         httpServer = new HddsDatanodeHttpServer(conf);
         httpServer.start();
         HttpConfig.Policy policy = HttpConfig.getHttpPolicy(conf);
+
         if (policy.isHttpEnabled()) {
-          datanodeDetails.setPort(DatanodeDetails.newPort(HTTP,
-                  httpServer.getHttpAddress().getPort()));
+          int httpPort = httpServer.getHttpAddress().getPort();
+          datanodeDetails.setPort(DatanodeDetails.newPort(HTTP, httpPort));
+          serviceRuntimeInfo.setHttpPort(String.valueOf(httpPort));
         }
+
         if (policy.isHttpsEnabled()) {
-          datanodeDetails.setPort(DatanodeDetails.newPort(HTTPS,
-                  httpServer.getHttpsAddress().getPort()));
+          int httpsPort = httpServer.getHttpAddress().getPort();
+          datanodeDetails.setPort(DatanodeDetails.newPort(HTTPS, httpsPort));
+          serviceRuntimeInfo.setHttpsPort(String.valueOf(httpsPort));
         }
+
       } catch (Exception ex) {
         LOG.error("HttpServer failed to start.", ex);
       }
 
-
       clientProtocolServer = new HddsDatanodeClientProtocolServer(
           datanodeDetails, conf, HddsVersionInfo.HDDS_VERSION_INFO,
           reconfigurationHandler);
 
+      int clientRpcport = clientProtocolServer.getClientRpcAddress().getPort();
+      serviceRuntimeInfo.setClientRpcPort(String.valueOf(clientRpcport));
+
       // Get admin list
       String starterUser =
           UserGroupInformation.getCurrentUser().getShortUserName();

hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/common/impl/ContainerSet.java

@@ -251,6 +251,21 @@ public Iterator<Container<?>> getContainerIterator(HddsVolume volume) {
         .iterator();
   }
 
+  /**
+   * Get the number of containers based on the given volume.
+   *
+   * @param volume hdds volume.
+   * @return number of containers
+   */
+  public long containerCount(HddsVolume volume) {
+    Preconditions.checkNotNull(volume);
+    Preconditions.checkNotNull(volume.getStorageID());
+    String volumeUuid = volume.getStorageID();
+    return containerMap.values().stream()
+        .filter(x -> volumeUuid.equals(x.getContainerData().getVolume()
+        .getStorageID())).count();
+  }
+
   /**
    * Return an containerMap iterator over {@link ContainerSet#containerMap}.
    * @return containerMap Iterator

hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/common/volume/HddsVolume.java

@@ -29,13 +29,15 @@
 import org.apache.commons.io.FileUtils;
 import org.apache.hadoop.hdds.annotation.InterfaceAudience;
 import org.apache.hadoop.hdds.annotation.InterfaceStability;
+import org.apache.hadoop.hdds.conf.ConfigurationSource;
 import org.apache.hadoop.hdds.upgrade.HDDSLayoutFeature;
 import org.apache.hadoop.hdfs.server.datanode.checker.VolumeCheckResult;
 import org.apache.hadoop.ozone.container.common.statemachine.DatanodeConfiguration;
 import org.apache.hadoop.ozone.container.common.utils.DatanodeStoreCache;
 import org.apache.hadoop.ozone.container.common.utils.HddsVolumeUtil;
 import org.apache.hadoop.ozone.container.common.utils.RawDB;
 import org.apache.hadoop.ozone.container.common.utils.StorageVolumeUtil;
+import org.apache.hadoop.ozone.container.ozoneimpl.ContainerController;
 import org.apache.hadoop.ozone.container.upgrade.VersionedDatanodeFeatures;
 import org.apache.hadoop.ozone.container.upgrade.VersionedDatanodeFeatures.SchemaV3;
 import org.apache.hadoop.util.Time;
@@ -44,6 +46,7 @@
 
 import jakarta.annotation.Nullable;
 
+import static org.apache.hadoop.hdds.HddsConfigKeys.OZONE_DATANODE_IO_METRICS_PERCENTILES_INTERVALS_SECONDS_KEY;
 import static org.apache.hadoop.ozone.OzoneConsts.CONTAINER_DB_NAME;
 import static org.apache.hadoop.ozone.container.common.utils.HddsVolumeUtil.initPerDiskDBStore;
 
@@ -80,6 +83,8 @@ public class HddsVolume extends StorageVolume {
   private final VolumeIOStats volumeIOStats;
   private final VolumeInfoMetrics volumeInfoMetrics;
 
+  private ContainerController controller;
+
   private final AtomicLong committedBytes = new AtomicLong(); // till Open containers become full
 
   // Mentions the type of volume
@@ -119,8 +124,10 @@ private HddsVolume(Builder b) throws IOException {
 
     if (!b.getFailedVolume() && getVolumeInfo().isPresent()) {
       this.setState(VolumeState.NOT_INITIALIZED);
+      ConfigurationSource conf = getConf();
+      int[] intervals = conf.getInts(OZONE_DATANODE_IO_METRICS_PERCENTILES_INTERVALS_SECONDS_KEY);
       this.volumeIOStats = new VolumeIOStats(b.getVolumeRootStr(),
-          this.getStorageDir().toString());
+          this.getStorageDir().toString(), intervals);
       this.volumeInfoMetrics =
           new VolumeInfoMetrics(b.getVolumeRootStr(), this);
 
@@ -382,6 +389,17 @@ public void loadDbStore(boolean readOnly) throws IOException {
         getStorageID());
   }
 
+  public void setController(ContainerController controller) {
+    this.controller = controller;
+  }
+
+  public long getContainers() {
+    if (controller != null) {
+      return controller.getContainerCount(this);
+    }
+    return 0;
+  }
+
   /**
    * Pick a DbVolume for HddsVolume and init db instance.
    * Use the HddsVolume directly if no DbVolume found.

hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/common/volume/VolumeIOStats.java

@@ -21,20 +21,34 @@
 import org.apache.hadoop.metrics2.MetricsSystem;
 import org.apache.hadoop.metrics2.annotation.Metric;
 import org.apache.hadoop.metrics2.lib.DefaultMetricsSystem;
+import org.apache.hadoop.metrics2.lib.MetricsRegistry;
 import org.apache.hadoop.metrics2.lib.MutableCounterLong;
+import org.apache.hadoop.metrics2.lib.MutableQuantiles;
+import org.apache.hadoop.metrics2.lib.MutableRate;
 
 /**
  * This class is used to track Volume IO stats for each HDDS Volume.
  */
 public class VolumeIOStats {
   private String metricsSourceName = VolumeIOStats.class.getSimpleName();
   private String storageDirectory;
-  private @Metric MutableCounterLong readBytes;
-  private @Metric MutableCounterLong readOpCount;
-  private @Metric MutableCounterLong writeBytes;
-  private @Metric MutableCounterLong writeOpCount;
-  private @Metric MutableCounterLong readTime;
-  private @Metric MutableCounterLong writeTime;
+  private final MetricsRegistry registry = new MetricsRegistry("VolumeIOStats");
+  @Metric
+  private MutableCounterLong readBytes;
+  @Metric
+  private MutableCounterLong readOpCount;
+  @Metric
+  private MutableCounterLong writeBytes;
+  @Metric
+  private MutableCounterLong writeOpCount;
+  @Metric
+  private MutableRate readTime;
+  @Metric
+  private MutableQuantiles[] readLatencyQuantiles;
+  @Metric
+  private MutableRate writeTime;
+  @Metric
+  private MutableQuantiles[] writeLatencyQuantiles;
 
   @Deprecated
   public VolumeIOStats() {
@@ -44,9 +58,24 @@ public VolumeIOStats() {
   /**
    * @param identifier Typically, path to volume root. e.g. /data/hdds
    */
-  public VolumeIOStats(String identifier, String storageDirectory) {
+  public VolumeIOStats(String identifier, String storageDirectory, int[] intervals) {
     this.metricsSourceName += '-' + identifier;
     this.storageDirectory = storageDirectory;
+
+    // Try initializing `readLatencyQuantiles` and `writeLatencyQuantiles`
+    if (intervals != null && intervals.length > 0) {
+      final int length = intervals.length;
+      readLatencyQuantiles = new MutableQuantiles[intervals.length];
+      writeLatencyQuantiles = new MutableQuantiles[intervals.length];
+      for (int i = 0; i < length; i++) {
+        readLatencyQuantiles[i] = registry.newQuantiles(
+            "readLatency" + intervals[i] + "s",
+            "Read Data File Io Latency in ms", "ops", "latency", intervals[i]);
+        writeLatencyQuantiles[i] = registry.newQuantiles(
+            "writeLatency" + intervals[i] + "s",
+            "Write Data File Io Latency in ms", "ops", "latency", intervals[i]);
+      }
+    }
     init();
   }
 
@@ -99,15 +128,21 @@ public void incWriteOpCount() {
    * @param time
    */
   public void incReadTime(long time) {
-    readTime.incr(time);
+    readTime.add(time);
+    for (MutableQuantiles q : readLatencyQuantiles) {
+      q.add(time);
+    }
   }
 
   /**
    * Increment the time taken by write operation on the volume.
    * @param time
    */
   public void incWriteTime(long time) {
-    writeTime.incr(time);
+    writeTime.add(time);
+    for (MutableQuantiles q : writeLatencyQuantiles) {
+      q.add(time);
+    }
   }
 
   /**
@@ -147,15 +182,15 @@ public long getWriteOpCount() {
    * @return long
    */
   public long getReadTime() {
-    return readTime.value();
+    return (long) readTime.lastStat().total();
   }
 
   /**
    * Returns total write operations time on the volume.
    * @return long
    */
   public long getWriteTime() {
-    return writeTime.value();
+    return (long) writeTime.lastStat().total();
   }
 
   @Metric

hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/common/volume/VolumeInfoMetrics.java

@@ -37,6 +37,7 @@ public class VolumeInfoMetrics {
   private final HddsVolume volume;
   @Metric("Returns the RocksDB compact times of the Volume")
   private MutableRate dbCompactLatency;
+  private long containers;
 
   /**
    * @param identifier Typically, path to volume root. E.g. /data/hdds
@@ -153,4 +154,11 @@ public void dbCompactTimesNanoSecondsIncr(long time) {
     dbCompactLatency.add(time);
   }
 
+  /**
+   * Return the Container Count of the Volume.
+   */
+  @Metric("Returns the Container Count of the Volume")
+  public long getContainers() {
+    return volume.getContainers();
+  }
 }

hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/ozoneimpl/BackgroundContainerDataScanner.java

@@ -63,6 +63,7 @@ public BackgroundContainerDataScanner(ContainerScannerConfiguration conf,
     throttler = new HddsDataTransferThrottler(conf.getBandwidthPerVolume());
     canceler = new Canceler();
     this.metrics = ContainerDataScannerMetrics.create(volume.toString());
+    this.metrics.setStorageDirectory(volume.toString());
     this.minScanGap = conf.getContainerScanMinGap();
   }