Logging fileSize of log files under NM Local Dir. Contributed by Prabhu Joseph

(cherry picked from commit 54ac80176e)

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

@@ -1322,6 +1322,11 @@ public class YarnConfiguration extends Configuration {
   public static final String LOG_AGGREGATION_RETAIN_SECONDS = YARN_PREFIX
       + "log-aggregation.retain-seconds";
   public static final long DEFAULT_LOG_AGGREGATION_RETAIN_SECONDS = -1;
+
+  public static final String LOG_AGGREGATION_DEBUG_FILESIZE = YARN_PREFIX
+      + "log-aggregation.debug.filesize";
+  public static final long DEFAULT_LOG_AGGREGATION_DEBUG_FILESIZE
+      = 100 * 1024 * 1024;
   
   /**
    * How long to wait between aggregated log retention checks. If set to

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

@@ -1262,6 +1262,14 @@
     <value>-1</value>
   </property>
 
+  <property>
+    <description>The log files created under NM Local Directories
+    will be logged if it exceeds the configured bytes. This
+    only takes effect if log4j level is at least Debug.</description>
+    <name>yarn.log-aggregation.debug.filesize</name>
+    <value>104857600</value>
+  </property>
+
   <property>
     <description>Specify which log file controllers we will support. The first
     file controller we add will be used to write the aggregated logs.

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/containermanager/logaggregation/AppLogAggregatorImpl.java

@@ -108,7 +108,7 @@ public class AppLogAggregatorImpl implements AppLogAggregator {
   private final AtomicBoolean waiting = new AtomicBoolean(false);
   private int logAggregationTimes = 0;
   private int cleanupOldLogTimes = 0;
-
+  private long logFileSizeThreshold;
   private boolean renameTemporaryLogFileFailed = false;
 
   private final Map<ContainerId, ContainerLogAggregator> containerLogAggregators =
@@ -175,6 +175,9 @@ public class AppLogAggregatorImpl implements AppLogAggregator {
     this.nodeId = nodeId;
     this.logAggPolicy = getLogAggPolicy(conf);
     this.recoveredLogInitedTime = recoveredLogInitedTime;
+    this.logFileSizeThreshold =
+        conf.getLong(YarnConfiguration.LOG_AGGREGATION_DEBUG_FILESIZE,
+        YarnConfiguration.DEFAULT_LOG_AGGREGATION_DEBUG_FILESIZE);
     if (logAggregationFileController == null) {
       // by default, use T-File Controller
       this.logAggregationFileController = new LogAggregationTFileController();
@@ -327,6 +330,19 @@ public class AppLogAggregatorImpl implements AppLogAggregator {
           uploadedLogsInThisCycle = true;
           List<Path> uploadedFilePathsInThisCycleList = new ArrayList<>();
           uploadedFilePathsInThisCycleList.addAll(uploadedFilePathsInThisCycle);
+          if (LOG.isDebugEnabled()) {
+            for (Path uploadedFilePath : uploadedFilePathsInThisCycleList) {
+              try {
+                long fileSize = lfs.getFileStatus(uploadedFilePath).getLen();
+                if (fileSize >= logFileSizeThreshold) {
+                  LOG.debug("Log File " + uploadedFilePath
+                      + " size is " + fileSize + " bytes");
+                }
+              } catch (Exception e1) {
+                LOG.error("Failed to get log file size " + e1);
+              }
+            }
+          }
           DeletionTask deletionTask = new FileDeletionTask(delService,
               this.userUgi.getShortUserName(), null,
               uploadedFilePathsInThisCycleList);