YARN-4309. Add container launch related debug information to container logs when a container fails. (Varun Vasudev via wangda)

(cherry picked from commit dfcbbdd)

Author: wangdatan

hadoop-yarn-project/CHANGES.txt

@@ -561,6 +561,9 @@ Release 2.8.0 - UNRELEASED
     YARN-3946. Update exact reason as to why a submitted app is in ACCEPTED state to 
     app's diagnostic message. (Naganarasimha G R via wangda)
 
+    YARN-4309. Add container launch related debug information to container logs
+    when a container fails. (Varun Vasudev via wangda)
+
   OPTIMIZATIONS
 
     YARN-3339. TestDockerContainerExecutor should pull a single image and not

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

@@ -92,6 +92,11 @@ private static void addDeprecatedKeys() {
   /** Delay before deleting resource to ease debugging of NM issues */
   public static final String DEBUG_NM_DELETE_DELAY_SEC =
     YarnConfiguration.NM_PREFIX + "delete.debug-delay-sec";
+
+  public static final String NM_LOG_CONTAINER_DEBUG_INFO =
+      YarnConfiguration.NM_PREFIX + "log-container-debug-info.enabled";
+
+  public static final boolean DEFAULT_NM_LOG_CONTAINER_DEBUG_INFO = false;
 
   ////////////////////////////////
   // IPC Configs

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

@@ -1127,6 +1127,18 @@
     <value>logs</value>
   </property>
 
+  <property>
+    <description>Generate additional logs about container launches.
+    Currently, this creates a copy of the launch script and lists the
+    directory contents of the container work dir. When listing directory
+    contents, we follow symlinks to a max-depth of 5(including symlinks
+    which point to outside the container work dir) which may lead to a
+    slowness in launching containers.
+    </description>
+    <name>yarn.nodemanager.log-container-debug-info.enabled</name>
+    <value>false</value>
+  </property>
+
   <property>
     <description>Amount of physical memory, in MB, that can be allocated 
     for containers. If set to -1 and

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/ContainerExecutor.java

@@ -34,6 +34,7 @@
 import java.util.concurrent.locks.ReentrantReadWriteLock.ReadLock;
 import java.util.concurrent.locks.ReentrantReadWriteLock.WriteLock;
 
+import com.google.common.annotations.VisibleForTesting;
 import org.apache.commons.io.FileUtils;
 import org.apache.commons.logging.Log;
 import org.apache.commons.logging.LogFactory;
@@ -65,6 +66,8 @@ public abstract class ContainerExecutor implements Configurable {
   final public static FsPermission TASK_LAUNCH_SCRIPT_PERMISSION =
     FsPermission.createImmutable((short) 0700);
 
+  public static final String DIRECTORY_CONTENTS = "directory.info";
+
   private Configuration conf;
 
   private ConcurrentMap<ContainerId, Path> pidFiles =
@@ -241,11 +244,22 @@ public int reacquireContainer(ContainerReacquisitionContext ctx)
    * @param resources The resources which have been localized for this container
    * Symlinks will be created to these localized resources
    * @param command The command that will be run.
+   * @param logDir The log dir to copy debugging information to
    * @throws IOException if any errors happened writing to the OutputStream,
    * while creating symlinks
    */
   public void writeLaunchEnv(OutputStream out, Map<String, String> environment,
-    Map<Path, List<String>> resources, List<String> command) throws IOException{
+      Map<Path, List<String>> resources, List<String> command, Path logDir)
+      throws IOException {
+    this.writeLaunchEnv(out, environment, resources, command, logDir,
+        ContainerLaunch.CONTAINER_SCRIPT);
+  }
+
+  @VisibleForTesting
+  public void writeLaunchEnv(OutputStream out,
+      Map<String, String> environment, Map<Path, List<String>> resources,
+      List<String> command, Path logDir, String outFilename)
+      throws IOException {
     ContainerLaunch.ShellScriptBuilder sb =
       ContainerLaunch.ShellScriptBuilder.create();
     Set<String> whitelist = new HashSet<String>();
@@ -272,6 +286,14 @@ public void writeLaunchEnv(OutputStream out, Map<String, String> environment,
       }
     }
 
+    // dump debugging information if configured
+    if (getConf() != null && getConf().getBoolean(
+        YarnConfiguration.NM_LOG_CONTAINER_DEBUG_INFO,
+        YarnConfiguration.DEFAULT_NM_LOG_CONTAINER_DEBUG_INFO)) {
+      sb.copyDebugInformation(new Path(outFilename), new Path(logDir, outFilename));
+      sb.listDebugInformation(new Path(logDir, DIRECTORY_CONTENTS));
+    }
+
     sb.command(command);
 
     PrintStream pout = null;

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/DockerContainerExecutor.java

@@ -329,7 +329,7 @@ public int launchContainer(ContainerStartContext ctx) throws IOException {
    * the docker image and write them out to an OutputStream.
    */
   public void writeLaunchEnv(OutputStream out, Map<String, String> environment,
-    Map<Path, List<String>> resources, List<String> command)
+    Map<Path, List<String>> resources, List<String> command, Path logDir)
     throws IOException {
     ContainerLaunch.ShellScriptBuilder sb =
       ContainerLaunch.ShellScriptBuilder.create();
@@ -358,6 +358,15 @@ public void writeLaunchEnv(OutputStream out, Map<String, String> environment,
       }
     }
 
+    // dump debugging information if configured
+    if (getConf() != null && getConf().getBoolean(
+        YarnConfiguration.NM_LOG_CONTAINER_DEBUG_INFO,
+        YarnConfiguration.DEFAULT_NM_LOG_CONTAINER_DEBUG_INFO)) {
+      sb.copyDebugInformation(new Path(ContainerLaunch.CONTAINER_SCRIPT),
+          new Path(logDir, ContainerLaunch.CONTAINER_SCRIPT));
+      sb.listDebugInformation(new Path(logDir, DIRECTORY_CONTENTS));
+    }
+
     sb.command(command);
 
     PrintStream pout = null;

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/containermanager/launcher/ContainerLaunch.java

@@ -272,7 +272,8 @@ public Integer call() {
 
         // Write out the environment
         exec.writeLaunchEnv(containerScriptOutStream, environment,
-          localResources, launchContext.getCommands());
+          localResources, launchContext.getCommands(),
+            new Path(containerLogDirs.get(0)));
 
         // /////////// End of writing out container-script
 
@@ -641,6 +642,28 @@ public final void symlink(Path src, Path dst) throws IOException {
       link(src, dst);
     }
 
+    /**
+     * Method to copy files that are useful for debugging container failures.
+     * This method will be called by ContainerExecutor when setting up the
+     * container launch script. The method should take care to make sure files
+     * are read-able by the yarn user if the files are to undergo
+     * log-aggregation.
+     * @param src path to the source file
+     * @param dst path to the destination file - should be absolute
+     * @throws IOException
+     */
+    public abstract void copyDebugInformation(Path src, Path dst)
+        throws IOException;
+
+    /**
+     * Method to dump debug information to the a target file. This method will
+     * be called by ContainerExecutor when setting up the container launch
+     * script.
+     * @param output the file to which debug information is to be written
+     * @throws IOException
+     */
+    public abstract void listDebugInformation(Path output) throws IOException;
+
     @Override
     public String toString() {
       return sb.toString();
@@ -704,6 +727,36 @@ protected void mkdir(Path path) {
       line("mkdir -p ", path.toString());
       errorCheck();
     }
+
+    @Override
+    public void copyDebugInformation(Path src, Path dest) throws IOException {
+      line("# Creating copy of launch script");
+      line("cp \"", src.toUri().getPath(), "\" \"", dest.toUri().getPath(),
+          "\"");
+      // set permissions to 640 because we need to be able to run
+      // log aggregation in secure mode as well
+      if(dest.isAbsolute()) {
+        line("chmod 640 \"", dest.toUri().getPath(), "\"");
+      }
+    }
+
+    @Override
+    public void listDebugInformation(Path output) throws  IOException {
+      line("# Determining directory contents");
+      line("echo \"ls -l:\" 1>\"", output.toString(), "\"");
+      line("ls -l 1>>\"", output.toString(), "\"");
+
+      // don't run error check because if there are loops
+      // find will exit with an error causing container launch to fail
+      // find will follow symlinks outside the work dir if such sylimks exist
+      // (like public/app local resources)
+      line("echo \"find -L . -maxdepth 5 -ls:\" 1>>\"", output.toString(),
+          "\"");
+      line("find -L . -maxdepth 5 -ls 1>>\"", output.toString(), "\"");
+      line("echo \"broken symlinks(find -L . -maxdepth 5 -type l -ls):\" 1>>\"",
+          output.toString(), "\"");
+      line("find -L . -maxdepth 5 -type l -ls 1>>\"", output.toString(), "\"");
+    }
   }
 
   private static final class WindowsShellScriptBuilder
@@ -757,6 +810,25 @@ protected void mkdir(Path path) throws IOException {
           path.toString(), path.toString()));
       errorCheck();
     }
+
+    @Override
+    public void copyDebugInformation(Path src, Path dest)
+        throws IOException {
+      // no need to worry about permissions - in secure mode
+      // WindowsSecureContainerExecutor will set permissions
+      // to allow NM to read the file
+      line("rem Creating copy of launch script");
+      lineWithLenCheck(String.format("copy \"%s\" \"%s\"", src.toString(),
+          dest.toString()));
+    }
+
+    @Override
+    public void listDebugInformation(Path output) throws IOException {
+      line("rem Determining directory contents");
+      lineWithLenCheck(
+          String.format("@echo \"dir:\" > \"%s\"", output.toString()));
+      lineWithLenCheck(String.format("dir >> \"%s\"", output.toString()));
+    }
   }
 
   private static void putEnvIfNotNull(