YARN-8155. Improve ATSv2 client logging in RM and NM publisher. Contributed by Abhishek Modi.

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

@@ -190,9 +190,20 @@ public class NMTimelinePublisher extends CompositeService {
           LOG.error("Seems like client has been removed before the container"
               + " metric could be published for " + container.getContainerId());
         }
-      } catch (IOException | YarnException e) {
+      } catch (IOException e) {
         LOG.error("Failed to publish Container metrics for container "
-            + container.getContainerId(), e);
+            + container.getContainerId());
+        if (LOG.isDebugEnabled()) {
+          LOG.debug("Failed to publish Container metrics for container "
+              + container.getContainerId(), e);
+        }
+      } catch (YarnException e) {
+        LOG.error("Failed to publish Container metrics for container "
+            + container.getContainerId(), e.getMessage());
+        if (LOG.isDebugEnabled()) {
+          LOG.debug("Failed to publish Container metrics for container "
+              + container.getContainerId(), e);
+        }
       }
     }
   }
@@ -284,9 +295,20 @@ public class NMTimelinePublisher extends CompositeService {
         LOG.error("Seems like client has been removed before the event could be"
             + " published for " + container.getContainerId());
       }
-    } catch (IOException | YarnException e) {
+    } catch (IOException e) {
+      LOG.error("Failed to publish Container metrics for container "
+          + container.getContainerId());
+      if (LOG.isDebugEnabled()) {
+        LOG.debug("Failed to publish Container metrics for container "
+            + container.getContainerId(), e);
+      }
+    } catch (YarnException e) {
       LOG.error("Failed to publish Container metrics for container "
-          + container.getContainerId(), e);
+          + container.getContainerId(), e.getMessage());
+      if (LOG.isDebugEnabled()) {
+        LOG.debug("Failed to publish Container metrics for container "
+            + container.getContainerId(), e);
+      }
     }
   }
 
@@ -315,8 +337,16 @@ public class NMTimelinePublisher extends CompositeService {
         LOG.error("Seems like client has been removed before the entity "
             + "could be published for " + entity);
       }
-    } catch (Exception e) {
-      LOG.error("Error when publishing entity " + entity, e);
+    } catch (IOException e) {
+      LOG.error("Error when publishing entity " + entity);
+      if (LOG.isDebugEnabled()) {
+        LOG.debug("Error when publishing entity " + entity, e);
+      }
+    } catch (YarnException e) {
+      LOG.error("Error when publishing entity " + entity, e.getMessage());
+      if (LOG.isDebugEnabled()) {
+        LOG.debug("Error when publishing entity " + entity, e);
+      }
     }
   }
 

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/metrics/TimelineServiceV2Publisher.java

@@ -18,6 +18,7 @@
 
 package org.apache.hadoop.yarn.server.resourcemanager.metrics;
 
+import java.io.IOException;
 import java.util.HashMap;
 import java.util.HashSet;
 import java.util.Map;
@@ -457,8 +458,11 @@ public class TimelineServiceV2Publisher extends AbstractSystemMetricsPublisher {
       entities.addEntity(entity);
       timelineCollector.putEntities(entities,
           UserGroupInformation.getCurrentUser());
-    } catch (Exception e) {
-      LOG.error("Error when publishing entity " + entity, e);
+    } catch (IOException e) {
+      LOG.error("Error when publishing entity " + entity);
+      if (LOG.isDebugEnabled()) {
+        LOG.debug("Error when publishing entity " + entity, e);
+      }
     }
   }
 

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-timelineservice/src/main/java/org/apache/hadoop/yarn/server/timelineservice/collector/TimelineCollectorWebService.java

@@ -61,6 +61,9 @@ import com.google.inject.Singleton;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
+import java.io.IOException;
+import java.util.IllegalFormatException;
+
 /**
  * The main per-node REST end point for timeline service writes. It is
  * essentially a container service that routes requests to the appropriate
@@ -166,7 +169,7 @@ public class TimelineCollectorWebService {
       TimelineCollector collector = collectorManager.get(appID);
       if (collector == null) {
         LOG.error("Application: "+ appId + " is not found");
-        throw new NotFoundException(); // different exception?
+        throw new NotFoundException("Application: "+ appId + " is not found");
       }
 
       boolean isAsync = async != null && async.trim().equalsIgnoreCase("true");
@@ -179,7 +182,10 @@ public class TimelineCollectorWebService {
       }
 
       return Response.ok().build();
-    } catch (Exception e) {
+    } catch (NotFoundException | ForbiddenException e) {
+      throw new WebApplicationException(e,
+          Response.Status.INTERNAL_SERVER_ERROR);
+    } catch (IOException e) {
       LOG.error("Error putting entities", e);
       throw new WebApplicationException(e,
           Response.Status.INTERNAL_SERVER_ERROR);
@@ -221,14 +227,17 @@ public class TimelineCollectorWebService {
       TimelineCollector collector = collectorManager.get(appID);
       if (collector == null) {
         LOG.error("Application: " + appId + " is not found");
-        throw new NotFoundException(); // different exception?
+        throw new NotFoundException("Application: " + appId + " is not found");
       }
 
       domain.setOwner(callerUgi.getShortUserName());
       collector.putDomain(domain, callerUgi);
 
       return Response.ok().build();
-    } catch (Exception e) {
+    } catch (NotFoundException e) {
+      throw new WebApplicationException(e,
+          Response.Status.INTERNAL_SERVER_ERROR);
+    } catch (IOException e) {
       LOG.error("Error putting entities", e);
       throw new WebApplicationException(e,
           Response.Status.INTERNAL_SERVER_ERROR);
@@ -242,7 +251,7 @@ public class TimelineCollectorWebService {
       } else {
         return null;
       }
-    } catch (Exception e) {
+    } catch (IllegalFormatException e) {
       LOG.error("Invalid application ID: " + appId);
       return null;
     }