[hadoop] branch trunk updated: HDFS-15808. Add metrics for FSNamesystem read/write lock hold long time. (#2668) Contributed by tomscut.

[sh...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

shv 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 9cb51bf  HDFS-15808. Add metrics for FSNamesystem read/write lock hold long time. (#2668) Contributed by tomscut.
9cb51bf is described below

commit 9cb51bf106802c78b1400fba9f1d1c7e772dd5e7
Author: tomscut <li...@bigo.sg>
AuthorDate: Mon Mar 1 16:35:12 2021 -0800

    HDFS-15808. Add metrics for FSNamesystem read/write lock hold long time. (#2668) Contributed by tomscut.
---
 .../hadoop/hdfs/server/namenode/FSNamesystem.java  | 14 +++++++++
 .../hdfs/server/namenode/FSNamesystemLock.java     | 34 ++++++++++++++++++++++
 2 files changed, 48 insertions(+)

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 22b4b92..ff03d7b 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
@@ -4830,6 +4830,20 @@ public class FSNamesystem implements Namesystem, FSNamesystemMBean,
     return fsLock.getQueueLength();
   }
 
+  @Metric(value = {"ReadLockLongHoldCount", "The number of time " +
+          "the read lock has been held for longer than the threshold"},
+          type = Metric.Type.COUNTER)
+  public long getNumOfReadLockLongHold() {
+    return fsLock.getNumOfReadLockLongHold();
+  }
+
+  @Metric(value = {"WriteLockLongHoldCount", "The number of time " +
+          "the write lock has been held for longer than the threshold"},
+          type = Metric.Type.COUNTER)
+  public long getNumOfWriteLockLongHold() {
+    return fsLock.getNumOfWriteLockLongHold();
+  }
+
   int getNumberOfDatanodes(DatanodeReportType type) {
     readLock();
     try {
diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystemLock.java b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystemLock.java
index c03cfd5..842c6b3 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystemLock.java
+++ b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystemLock.java
@@ -109,6 +109,16 @@ class FSNamesystemLock {
   private final AtomicReference<LockHeldInfo> longestReadLockHeldInfo =
       new AtomicReference<>(new LockHeldInfo());
   private LockHeldInfo longestWriteLockHeldInfo = new LockHeldInfo();
+  /**
+   * The number of time the read lock
+   * has been held longer than the threshold.
+   */
+  private final AtomicLong numReadLockLongHold = new AtomicLong(0);
+  /**
+   * The number of time the write lock
+   * has been held for longer than the threshold.
+   */
+  private final AtomicLong numWriteLockLongHold = new AtomicLong(0);
 
   @VisibleForTesting
   static final String OP_NAME_OTHER = "OTHER";
@@ -182,6 +192,7 @@ class FSNamesystemLock {
     final long readLockIntervalMs =
         TimeUnit.NANOSECONDS.toMillis(readLockIntervalNanos);
     if (needReport && readLockIntervalMs >= this.readLockReportingThresholdMs) {
+      numReadLockLongHold.incrementAndGet();
       String lockReportInfo = null;
       boolean done = false;
       while (!done) {
@@ -298,6 +309,7 @@ class FSNamesystemLock {
     LogAction logAction = LogThrottlingHelper.DO_NOT_LOG;
     if (needReport &&
         writeLockIntervalMs >= this.writeLockReportingThresholdMs) {
+      numWriteLockLongHold.incrementAndGet();
       if (longestWriteLockHeldInfo.getIntervalMs() <= writeLockIntervalMs) {
         String lockReportInfo = lockReportInfoSupplier != null ? " (" +
             lockReportInfoSupplier.get() + ")" : "";
@@ -363,6 +375,28 @@ class FSNamesystemLock {
   }
 
   /**
+   * Returns the number of time the read lock
+   * has been held longer than the threshold.
+   *
+   * @return long - Number of time the read lock
+   * has been held longer than the threshold
+   */
+  public long getNumOfReadLockLongHold() {
+    return numReadLockLongHold.get();
+  }
+
+  /**
+   * Returns the number of time the write lock
+   * has been held longer than the threshold.
+   *
+   * @return long - Number of time the write lock
+   * has been held longer than the threshold.
+   */
+  public long getNumOfWriteLockLongHold() {
+    return numWriteLockLongHold.get();
+  }
+
+  /**
    * Add the lock hold time for a recent operation to the metrics.
    * @param operationName Name of the operation for which to record the time
    * @param value Length of time the lock was held (nanoseconds)


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