[YARN-10687] Add option to disable/enable free disk space checking and percentage checking for full and not-full disks. Contributed by Qi Zhu.

2021-03-12 17:17:31 +00:00 · 2021-03-12 17:17:31 +00:00 · 5f067cf0f3
commit 5f067cf0f3
parent 5db4c0bf70
4 changed files with 185 additions and 4 deletions
--- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api/src/main/java/org/apache/hadoop/yarn/conf/YarnConfiguration.java
+++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api/src/main/java/org/apache/hadoop/yarn/conf/YarnConfiguration.java
@ -2027,6 +2027,8 @@ public static boolean isAclEnabled(Configuration conf) {
   * marked as offline. Values can range from 0.0 to 100.0. If the value is
   * greater than or equal to 100, NM will check for full disk. This applies to
   * nm-local-dirs and nm-log-dirs.
+   *
+   * This applies when disk-utilization-threshold.enabled is true.
   */
  public static final String NM_MAX_PER_DISK_UTILIZATION_PERCENTAGE =
      NM_DISK_HEALTH_CHECK_PREFIX + "max-disk-utilization-per-disk-percentage";
@ -2036,6 +2038,17 @@ public static boolean isAclEnabled(Configuration conf) {
  public static final float DEFAULT_NM_MAX_PER_DISK_UTILIZATION_PERCENTAGE =
      90.0F;

+  /**
+   * Enable/Disable the disk utilisation percentage
+   * threshold for disk health checker.
+   */
+  public static final String NM_DISK_UTILIZATION_THRESHOLD_ENABLED =
+      NM_DISK_HEALTH_CHECK_PREFIX +
+          "disk-utilization-threshold.enabled";
+
+  public static final
+      boolean DEFAULT_NM_DISK_UTILIZATION_THRESHOLD_ENABLED = true;
+
  /**
   * The low threshold percentage of disk space used when an offline disk is
   * marked as online. Values can range from 0.0 to 100.0. The value shouldn't
@ -2051,9 +2064,23 @@ public static boolean isAclEnabled(Configuration conf) {
  /**
   * The minimum space that must be available on a local dir for it to be used.
   * This applies to nm-local-dirs and nm-log-dirs.
+   *
+   * This applies when disk-free-space-threshold.enabled is true.
   */
  public static final String NM_MIN_PER_DISK_FREE_SPACE_MB =
      NM_DISK_HEALTH_CHECK_PREFIX + "min-free-space-per-disk-mb";
+
+  /**
+   * Enable/Disable the minimum disk free
+   * space threshold for disk health checker.
+   */
+  public static final String NM_DISK_FREE_SPACE_THRESHOLD_ENABLED =
+      NM_DISK_HEALTH_CHECK_PREFIX +
+          "disk-free-space-threshold.enabled";
+
+  public static final boolean
+      DEFAULT_NM_DISK_FREE_SPACE_THRESHOLD_ENABLED = true;
+
  /**
   * The minimum space that must be available on an offline
   * disk for it to be marked as online.  The value should not be less
--- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common/src/main/resources/yarn-default.xml
+++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common/src/main/resources/yarn-default.xml
@ -1809,12 +1809,27 @@
    <value>0.25</value>
  </property>

+  <property>
+    <description>Enable/Disable the disk utilisation percentage
+      threshold for disk health checker.</description>
+    <name>yarn.nodemanager.disk-health-checker.disk-utilization-threshold.enabled</name>
+    <value>true</value>
+  </property>
+
+  <property>
+    <description> Enable/Disable the minimum disk free
+      space threshold for disk health checker.</description>
+    <name>yarn.nodemanager.disk-health-checker.disk-free-space-threshold.enabled</name>
+    <value>true</value>
+  </property>
+
  <property>
    <description>The maximum percentage of disk space utilization allowed after 
    which a disk is marked as bad. Values can range from 0.0 to 100.0. 
    If the value is greater than or equal to 100, the nodemanager will check 
    for full disk. This applies to yarn.nodemanager.local-dirs and
-    yarn.nodemanager.log-dirs.</description>
+    yarn.nodemanager.log-dirs when
+      yarn.nodemanager.disk-health-checker.disk-utilization-threshold.enabled is true.</description>
    <name>yarn.nodemanager.disk-health-checker.max-disk-utilization-per-disk-percentage</name>
    <value>90.0</value>
  </property>
@ -1834,7 +1849,8 @@
    <description>The minimum space in megabytes that must be available on a disk for
    it to be used. If space on a disk falls below this threshold, it will be marked
    as bad. This applies to yarn.nodemanager.local-dirs and
-    yarn.nodemanager.log-dirs.</description>
+    yarn.nodemanager.log-dirs when
+      yarn.nodemanager.disk-health-checker.disk-free-space-threshold.enabled is true.</description>
    <name>yarn.nodemanager.disk-health-checker.min-free-space-per-disk-mb</name>
    <value>0</value>
  </property>
--- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/DirectoryCollection.java
+++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/DirectoryCollection.java
@ -59,6 +59,9 @@ public class DirectoryCollection {

  private final Configuration conf;
  private final DiskValidator diskValidator;
+
+  private boolean diskUtilizationThresholdEnabled;
+  private boolean diskFreeSpaceThresholdEnabled;
  /**
   * The enum defines disk failure type.
   */
@ -239,6 +242,17 @@ public DirectoryCollection(String[] dirs,
      throw new YarnRuntimeException(e);
    }

+    diskUtilizationThresholdEnabled = conf.
+        getBoolean(YarnConfiguration.
+                NM_DISK_UTILIZATION_THRESHOLD_ENABLED,
+            YarnConfiguration.
+                DEFAULT_NM_DISK_UTILIZATION_THRESHOLD_ENABLED);
+    diskFreeSpaceThresholdEnabled = conf.
+        getBoolean(YarnConfiguration.
+                NM_DISK_FREE_SPACE_THRESHOLD_ENABLED,
+            YarnConfiguration.
+                DEFAULT_NM_DISK_FREE_SPACE_THRESHOLD_ENABLED);
+
    localDirs = new ArrayList<>(Arrays.asList(dirs));
    errorDirs = new ArrayList<>();
    fullDirs = new ArrayList<>();
@ -520,7 +534,9 @@ Map<String, DiskErrorInformation> testDirs(List<String> dirs,
            diskUtilizationPercentageCutoffHigh : diskUtilizationPercentageCutoffLow;
        long diskFreeSpaceCutoff = goodDirs.contains(dir) ?
            diskFreeSpaceCutoffLow : diskFreeSpaceCutoffHigh;
-        if (isDiskUsageOverPercentageLimit(testDir,
+
+        if (diskUtilizationThresholdEnabled
+            && isDiskUsageOverPercentageLimit(testDir,
            diskUtilizationPercentageCutoff)) {
          msg =
              "used space above threshold of "
@ -529,7 +545,8 @@ Map<String, DiskErrorInformation> testDirs(List<String> dirs,
          ret.put(dir,
            new DiskErrorInformation(DiskErrorCause.DISK_FULL, msg));
          continue;
-        } else if (isDiskFreeSpaceUnderLimit(testDir, diskFreeSpaceCutoff)) {
+        } else if (diskFreeSpaceThresholdEnabled
+            && isDiskFreeSpaceUnderLimit(testDir, diskFreeSpaceCutoff)) {
          msg =
              "free space below limit of " + diskFreeSpaceCutoff
                  + "MB";
@ -613,6 +630,28 @@ long getDiskUtilizationSpaceCutoffHigh() {
    return diskFreeSpaceCutoffHigh;
  }

+  @VisibleForTesting
+  boolean getDiskUtilizationThresholdEnabled() {
+    return diskUtilizationThresholdEnabled;
+  }
+
+  @VisibleForTesting
+  boolean getDiskFreeSpaceThresholdEnabled() {
+    return diskFreeSpaceThresholdEnabled;
+  }
+
+  @VisibleForTesting
+  void setDiskUtilizationThresholdEnabled(boolean
+      utilizationEnabled) {
+    diskUtilizationThresholdEnabled = utilizationEnabled;
+  }
+
+  @VisibleForTesting
+  void setDiskFreeSpaceThresholdEnabled(boolean
+      freeSpaceEnabled) {
+    diskFreeSpaceThresholdEnabled = freeSpaceEnabled;
+  }
+
  public void setDiskUtilizationSpaceCutoff(long freeSpaceCutoff) {
    setDiskUtilizationSpaceCutoff(freeSpaceCutoff,
        freeSpaceCutoff);
--- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/test/java/org/apache/hadoop/yarn/server/nodemanager/TestDirectoryCollection.java
+++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/test/java/org/apache/hadoop/yarn/server/nodemanager/TestDirectoryCollection.java
@ -176,6 +176,105 @@ public void testDiskSpaceUtilizationLimit() throws IOException {
      dc.getGoodDirsDiskUtilizationPercentage());
  }

+  @Test
+  public void testDiskSpaceUtilizationThresholdEnabled() throws IOException {
+
+    String dirA = new File(testDir, "dirA").getPath();
+    String[] dirs = {dirA};
+    DirectoryCollection dc = new DirectoryCollection(dirs, 0.0F);
+
+    // Disable disk utilization threshold.
+    dc.setDiskUtilizationThresholdEnabled(false);
+    Assert.assertFalse(dc.getDiskUtilizationThresholdEnabled());
+
+    dc.checkDirs();
+    Assert.assertEquals(1, dc.getGoodDirs().size());
+    Assert.assertEquals(0, dc.getErroredDirs().size());
+    Assert.assertEquals(0, dc.getFailedDirs().size());
+    Assert.assertEquals(0, dc.getFullDirs().size());
+    Assert.assertNull(dc.getDirectoryErrorInfo(dirA));
+
+    // Enable disk utilization threshold.
+    dc.setDiskUtilizationThresholdEnabled(true);
+    Assert.assertTrue(dc.getDiskUtilizationThresholdEnabled());
+
+    dc.checkDirs();
+    Assert.assertEquals(0, dc.getGoodDirs().size());
+    Assert.assertEquals(0, dc.getErroredDirs().size());
+    Assert.assertEquals(1, dc.getFailedDirs().size());
+    Assert.assertEquals(1, dc.getFullDirs().size());
+    Assert.assertNotNull(dc.getDirectoryErrorInfo(dirA));
+    Assert.assertEquals(DirectoryCollection.DiskErrorCause.DISK_FULL,
+        dc.getDirectoryErrorInfo(dirA).cause);
+
+    // no good dirs
+    Assert.assertEquals(0,
+        dc.getGoodDirsDiskUtilizationPercentage());
+
+    dc = new DirectoryCollection(dirs, 100.0F);
+    int utilizedSpacePerc =
+        (int) ((testDir.getTotalSpace() - testDir.getUsableSpace()) * 100 /
+            testDir.getTotalSpace());
+    dc.checkDirs();
+    Assert.assertEquals(1, dc.getGoodDirs().size());
+    Assert.assertEquals(0, dc.getErroredDirs().size());
+    Assert.assertEquals(0, dc.getFailedDirs().size());
+    Assert.assertEquals(0, dc.getFullDirs().size());
+    Assert.assertNull(dc.getDirectoryErrorInfo(dirA));
+
+    Assert.assertEquals(utilizedSpacePerc,
+        dc.getGoodDirsDiskUtilizationPercentage());
+
+    dc = new DirectoryCollection(dirs,
+        testDir.getTotalSpace() / (1024 * 1024));
+
+    // Disable disk utilization threshold.
+    dc.setDiskUtilizationThresholdEnabled(false);
+    Assert.assertFalse(dc.getDiskUtilizationThresholdEnabled());
+
+    // Disable disk free space threshold.
+    dc.setDiskFreeSpaceThresholdEnabled(false);
+    Assert.assertFalse(dc.getDiskFreeSpaceThresholdEnabled());
+    dc.checkDirs();
+
+    Assert.assertEquals(1, dc.getGoodDirs().size());
+    Assert.assertEquals(0, dc.getErroredDirs().size());
+    Assert.assertEquals(0, dc.getFailedDirs().size());
+    Assert.assertEquals(0, dc.getFullDirs().size());
+    Assert.assertNull(dc.getDirectoryErrorInfo(dirA));
+
+    dc = new DirectoryCollection(dirs,
+        testDir.getTotalSpace() / (1024 * 1024));
+
+    // Enable disk free space threshold.
+    dc.setDiskFreeSpaceThresholdEnabled(true);
+    Assert.assertTrue(dc.getDiskFreeSpaceThresholdEnabled());
+
+    dc.checkDirs();
+
+    Assert.assertEquals(0, dc.getGoodDirs().size());
+    Assert.assertEquals(0, dc.getErroredDirs().size());
+    Assert.assertEquals(1, dc.getFailedDirs().size());
+    Assert.assertEquals(1, dc.getFullDirs().size());
+    Assert.assertNotNull(dc.getDirectoryErrorInfo(dirA));
+    // no good dirs
+    Assert.assertEquals(0, dc.getGoodDirsDiskUtilizationPercentage());
+
+    dc = new DirectoryCollection(dirs, 100.0F, 100.0F, 0);
+    utilizedSpacePerc =
+        (int)((testDir.getTotalSpace() - testDir.getUsableSpace()) * 100 /
+            testDir.getTotalSpace());
+    dc.checkDirs();
+    Assert.assertEquals(1, dc.getGoodDirs().size());
+    Assert.assertEquals(0, dc.getErroredDirs().size());
+    Assert.assertEquals(0, dc.getFailedDirs().size());
+    Assert.assertEquals(0, dc.getFullDirs().size());
+    Assert.assertNull(dc.getDirectoryErrorInfo(dirA));
+
+    Assert.assertEquals(utilizedSpacePerc,
+        dc.getGoodDirsDiskUtilizationPercentage());
+  }
+
  @Test
  public void testDiskLimitsCutoffSetters() throws IOException {