This is an automated email from the ASF dual-hosted git repository.
hexiaoqiao pushed a commit to branch trunk
in repository https://gitbox.apache.org/repos/asf/hadoop.git
The following commit(s) were added to refs/heads/trunk by this push:
new 1d09dcc614e HDFS-17117. Print reconstructionQueuesInitProgress
periodically when BlockManager processMisReplicatesAsync. (#5877). Contributed
by Haiyang Hu.
1d09dcc614e is described below
commit 1d09dcc614e0faf212c657b276406ba1d69dc175
Author: huhaiyang <huhaiyang...@126.com>
AuthorDate: Thu Jul 27 11:28:30 2023 +0800
HDFS-17117. Print reconstructionQueuesInitProgress periodically when
BlockManager processMisReplicatesAsync. (#5877). Contributed by Haiyang Hu.
Signed-off-by: He Xiaoqiao <hexiaoq...@apache.org>
---
.../federation/metrics/NamenodeBeanMetrics.java | 5 +++
.../hdfs/server/blockmanagement/BlockManager.java | 12 ++++--
.../hadoop/hdfs/server/namenode/FSNamesystem.java | 9 +++++
.../server/namenode/metrics/FSNamesystemMBean.java | 7 ++++
.../server/namenode/TestFSNamesystemMBean.java | 44 ++++++++++++++++++++++
5 files changed, 73 insertions(+), 4 deletions(-)
diff --git
a/hadoop-hdfs-project/hadoop-hdfs-rbf/src/main/java/org/apache/hadoop/hdfs/server/federation/metrics/NamenodeBeanMetrics.java
b/hadoop-hdfs-project/hadoop-hdfs-rbf/src/main/java/org/apache/hadoop/hdfs/server/federation/metrics/NamenodeBeanMetrics.java
index 0c629221463..d16543e2d86 100644
---
a/hadoop-hdfs-project/hadoop-hdfs-rbf/src/main/java/org/apache/hadoop/hdfs/server/federation/metrics/NamenodeBeanMetrics.java
+++
b/hadoop-hdfs-project/hadoop-hdfs-rbf/src/main/java/org/apache/hadoop/hdfs/server/federation/metrics/NamenodeBeanMetrics.java
@@ -884,6 +884,11 @@ public class NamenodeBeanMetrics
return 0;
}
+ @Override
+ public float getReconstructionQueuesInitProgress() {
+ return 0;
+ }
+
private Router getRouter() throws IOException {
if (this.router == null) {
throw new IOException("Router is not initialized");
diff --git
a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockManager.java
b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockManager.java
index e09fe7fbd9b..d00bface655 100644
---
a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockManager.java
+++
b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockManager.java
@@ -449,7 +449,7 @@ public class BlockManager implements BlockStatsMXBean {
/**
* Progress of the Reconstruction queues initialisation.
*/
- private double reconstructionQueuesInitProgress = 0.0;
+ private float reconstructionQueuesInitProgress = 0.0f;
/** for block replicas placement */
private volatile BlockPlacementPolicies placementPolicies;
@@ -3889,8 +3889,10 @@ public class BlockManager implements BlockStatsMXBean {
totalProcessed += processed;
// there is a possibility that if any of the blocks deleted/added
during
// initialisation, then progress might be different.
- reconstructionQueuesInitProgress = Math.min((double) totalProcessed
- / totalBlocks, 1.0);
+ if (totalBlocks > 0) { // here avoid metrics appear as NaN.
+ reconstructionQueuesInitProgress = Math.min((float) totalProcessed
+ / totalBlocks, 1.0f);
+ }
if (!blocksItr.hasNext()) {
LOG.info("Total number of blocks = {}", blocksMap.size());
@@ -3910,6 +3912,8 @@ public class BlockManager implements BlockStatsMXBean {
}
} finally {
namesystem.writeUnlock("processMisReplicatesAsync");
+ LOG.info("Reconstruction queues initialisation progress: {}, total
number of blocks " +
+ "processed: {}/{}", reconstructionQueuesInitProgress,
totalProcessed, totalBlocks);
// Make sure it is out of the write lock for sufficiently long time.
Thread.sleep(sleepDuration);
}
@@ -3924,7 +3928,7 @@ public class BlockManager implements BlockStatsMXBean {
*
* @return Returns values between 0 and 1 for the progress.
*/
- public double getReconstructionQueuesInitProgress() {
+ public float getReconstructionQueuesInitProgress() {
return reconstructionQueuesInitProgress;
}
diff --git
a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java
b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java
index 87cfd172d1f..13a9809627e 100644
---
a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java
+++
b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java
@@ -4888,6 +4888,15 @@ public class FSNamesystem implements Namesystem,
FSNamesystemMBean,
return blockManager.getPendingSPSPaths();
}
+ /**
+ * Get the progress of the reconstruction queues initialisation.
+ */
+ @Override // FSNamesystemMBean
+ @Metric
+ public float getReconstructionQueuesInitProgress() {
+ return blockManager.getReconstructionQueuesInitProgress();
+ }
+
/**
* Returns the length of the wait Queue for the FSNameSystemLock.
*
diff --git
a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/metrics/FSNamesystemMBean.java
b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/metrics/FSNamesystemMBean.java
index 59bb01f0063..746cae7113a 100644
---
a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/metrics/FSNamesystemMBean.java
+++
b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/metrics/FSNamesystemMBean.java
@@ -261,4 +261,11 @@ public interface FSNamesystemMBean {
* @return The number of paths to be processed by sps.
*/
int getPendingSPSPaths();
+
+ /**
+ * Get the progress of the reconstruction queues initialisation.
+ *
+ * @return Returns values between 0 and 1 for the progress.
+ */
+ float getReconstructionQueuesInitProgress();
}
diff --git
a/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestFSNamesystemMBean.java
b/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestFSNamesystemMBean.java
index 5a7da057219..2353dd975a7 100644
---
a/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestFSNamesystemMBean.java
+++
b/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestFSNamesystemMBean.java
@@ -17,6 +17,7 @@
*/
package org.apache.hadoop.hdfs.server.namenode;
+import static org.junit.Assert.assertEquals;
import static org.junit.Assert.assertTrue;
import static org.junit.Assert.assertNotNull;
@@ -33,9 +34,12 @@ import javax.management.ObjectName;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.hdfs.DFSTestUtil;
+import org.apache.hadoop.hdfs.DistributedFileSystem;
import org.apache.hadoop.hdfs.MiniDFSCluster;
import org.apache.hadoop.metrics2.impl.ConfigBuilder;
import org.apache.hadoop.metrics2.impl.TestMetricsConfig;
+import org.apache.hadoop.test.GenericTestUtils;
import org.junit.Test;
import org.eclipse.jetty.util.ajax.JSON;
@@ -225,4 +229,44 @@ public class TestFSNamesystemMBean {
}
}
}
+
+ /**
+ * Test metrics associated with reconstructionQueuesInitProgress.
+ */
+ @Test
+ public void testReconstructionQueuesInitProgressMetrics() throws Exception {
+ Configuration conf = new Configuration();
+ try (MiniDFSCluster cluster = new MiniDFSCluster.Builder(conf).build()) {
+ cluster.waitActive();
+ final FSNamesystem fsNamesystem = cluster.getNamesystem();
+ final DistributedFileSystem fs = cluster.getFileSystem();
+
+ // Validate init reconstructionQueuesInitProgress value.
+ assertEquals(0.0, fsNamesystem.getReconstructionQueuesInitProgress(), 0);
+ MBeanServer mbs = ManagementFactory.getPlatformMBeanServer();
+ ObjectName mxbeanName =
+ new ObjectName("Hadoop:service=NameNode,name=FSNamesystemState");
+ float reconstructionQueuesInitProgress =
+ (float) mbs.getAttribute(mxbeanName,
"ReconstructionQueuesInitProgress");
+ assertEquals(0.0, reconstructionQueuesInitProgress, 0);
+
+ // Create file.
+ Path file = new Path("/test");
+ long fileLength = 1024 * 1024 * 3;
+ DFSTestUtil.createFile(fs, file, fileLength, (short) 1, 0L);
+ DFSTestUtil.waitReplication(fs, file, (short) 1);
+
+ // Restart nameNode to run processMisReplicatedBlocks.
+ cluster.restartNameNode(true);
+
+ // Validate reconstructionQueuesInitProgress value.
+ GenericTestUtils.waitFor(
+ () -> cluster.getNamesystem().getReconstructionQueuesInitProgress()
== 1.0,
+ 100, 5 * 1000);
+
+ reconstructionQueuesInitProgress =
+ (float) mbs.getAttribute(mxbeanName,
"ReconstructionQueuesInitProgress");
+ assertEquals(1.0, reconstructionQueuesInitProgress, 0);
+ }
+ }
}
---------------------------------------------------------------------
To unsubscribe, e-mail: common-commits-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-commits-h...@hadoop.apache.org
