YARN-7667. Docker Stop grace period should be configurable. Contributed by Eric Badger

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api/src/main/java/org/apache/hadoop/yarn/conf/YarnConfiguration.java

@@ -1951,6 +1951,20 @@ public class YarnConfiguration extends Configuration {
    */
   public static final boolean DEFAULT_NM_DOCKER_ALLOW_DELAYED_REMOVAL = false;
 
+  /**
+   * A configurable value to pass to the Docker Stop command. This value
+   * defines the number of seconds between the docker stop command sending
+   * a SIGTERM and a SIGKILL.
+   */
+  public static final String NM_DOCKER_STOP_GRACE_PERIOD =
+      DOCKER_CONTAINER_RUNTIME_PREFIX + "stop.grace-period";
+
+  /**
+   * The default value for the grace period between the SIGTERM and the
+   * SIGKILL in the Docker Stop command.
+   */
+  public static final int DEFAULT_NM_DOCKER_STOP_GRACE_PERIOD = 10;
+
   /** The mode in which the Java Container Sandbox should run detailed by
    *  the JavaSandboxLinuxContainerRuntime. */
   public static final String YARN_CONTAINER_SANDBOX =

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common/src/main/resources/yarn-default.xml

@@ -1786,6 +1786,14 @@
     <value>false</value>
   </property>
 
+  <property>
+    <description>A configurable value to pass to the Docker Stop command. This value
+      defines the number of seconds between the docker stop command sending
+      a SIGTERM and a SIGKILL.</description>
+    <name>yarn.nodemanager.runtime.linux.docker.stop.grace-period</name>
+    <value>10</value>
+  </property>
+
   <property>
     <description>The mode in which the Java Container Sandbox should run detailed by
       the JavaSandboxLinuxContainerRuntime.</description>

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/containermanager/linux/runtime/DockerLinuxContainerRuntime.java

@@ -245,6 +245,7 @@ public class DockerLinuxContainerRuntime implements LinuxContainerRuntime {
   private int userRemappingGidThreshold;
   private Set<String> capabilities;
   private boolean delayedRemovalAllowed;
+  private int dockerStopGracePeriod;
 
   /**
    * Return whether the given environment variables indicate that the operation
@@ -348,6 +349,10 @@ public class DockerLinuxContainerRuntime implements LinuxContainerRuntime {
     delayedRemovalAllowed = conf.getBoolean(
         YarnConfiguration.NM_DOCKER_ALLOW_DELAYED_REMOVAL,
         YarnConfiguration.DEFAULT_NM_DOCKER_ALLOW_DELAYED_REMOVAL);
+
+    dockerStopGracePeriod = conf.getInt(
+        YarnConfiguration.NM_DOCKER_STOP_GRACE_PERIOD,
+        YarnConfiguration.DEFAULT_NM_DOCKER_STOP_GRACE_PERIOD);
   }
 
   private Set<String> getDockerCapabilitiesFromConf() throws
@@ -1138,7 +1143,8 @@ public class DockerLinuxContainerRuntime implements LinuxContainerRuntime {
         DockerCommandExecutor.getContainerStatus(containerId, conf,
             privilegedOperationExecutor);
     if (DockerCommandExecutor.isStoppable(containerStatus)) {
-      DockerStopCommand dockerStopCommand = new DockerStopCommand(containerId);
+      DockerStopCommand dockerStopCommand = new DockerStopCommand(
+          containerId).setGracePeriod(dockerStopGracePeriod);
       DockerCommandExecutor.executeDockerCommand(dockerStopCommand, containerId,
           env, conf, privilegedOperationExecutor, false);
     } else {

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/test/java/org/apache/hadoop/yarn/server/nodemanager/containermanager/linux/runtime/TestDockerContainerRuntime.java

@@ -146,6 +146,7 @@ public class TestDockerContainerRuntime {
   private final String whitelistedUser = "yoda";
   private String[] testCapabilities;
   private final String signalPid = "1234";
+  private int dockerStopGracePeriod;
 
   @Before
   public void setup() {
@@ -166,6 +167,10 @@ public class TestDockerContainerRuntime {
     env.put("FROM_CLIENT", "1");
     image = "busybox:latest";
 
+    dockerStopGracePeriod = conf.getInt(
+      YarnConfiguration.NM_DOCKER_STOP_GRACE_PERIOD,
+      YarnConfiguration.DEFAULT_NM_DOCKER_STOP_GRACE_PERIOD);
+
     env.put(DockerLinuxContainerRuntime.ENV_DOCKER_CONTAINER_IMAGE, image);
     when(container.getContainerId()).thenReturn(cId);
     when(cId.toString()).thenReturn(containerId);
@@ -1308,10 +1313,11 @@ public class TestDockerContainerRuntime {
     List<String> dockerCommands = getDockerCommandsForSignal(
         ContainerExecutor.Signal.TERM,
         DockerCommandExecutor.DockerContainerStatus.RUNNING);
-    Assert.assertEquals(3, dockerCommands.size());
+    Assert.assertEquals(4, dockerCommands.size());
     Assert.assertEquals("[docker-command-execution]", dockerCommands.get(0));
     Assert.assertEquals("  docker-command=stop", dockerCommands.get(1));
     Assert.assertEquals("  name=container_id", dockerCommands.get(2));
+    Assert.assertEquals("  time=10", dockerCommands.get(3));
   }
 
   @Test
@@ -1321,10 +1327,11 @@ public class TestDockerContainerRuntime {
     List<String> dockerCommands = getDockerCommandsForSignal(
         ContainerExecutor.Signal.KILL,
         DockerCommandExecutor.DockerContainerStatus.RUNNING);
-    Assert.assertEquals(3, dockerCommands.size());
+    Assert.assertEquals(4, dockerCommands.size());
     Assert.assertEquals("[docker-command-execution]", dockerCommands.get(0));
     Assert.assertEquals("  docker-command=stop", dockerCommands.get(1));
     Assert.assertEquals("  name=container_id", dockerCommands.get(2));
+    Assert.assertEquals("  time=10", dockerCommands.get(3));
   }
 
   @Test
@@ -1884,7 +1891,8 @@ public class TestDockerContainerRuntime {
             || ContainerExecutor.Signal.TERM.equals(signal)) {
           if (DockerCommandExecutor.isStoppable(containerStatus)) {
             DockerStopCommand dockerStopCommand =
-                new DockerStopCommand(containerName);
+                new DockerStopCommand(containerName)
+                .setGracePeriod(dockerStopGracePeriod);
             DockerCommandExecutor.executeDockerCommand(dockerStopCommand,
                 containerName, environment, conf, mockExecutor, false);
           }