YARN-8459. Improve Capacity Scheduler logs to debug invalid states. Contributed by Wangda Tan.

(cherry picked from commit 51654a3962bcd0482c0e1dd32765e9533e3f9158)

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/CapacityScheduler.java

@@ -1197,8 +1197,10 @@ public class CapacityScheduler extends
       updateDemandForQueue.getOrderingPolicy().demandUpdated(application);
     }
 
-    LOG.info("Allocation for application " + applicationAttemptId + " : " +
-        allocation + " with cluster resource : " + getClusterResource());
+    if (LOG.isDebugEnabled()) {
+      LOG.info("Allocation for application " + applicationAttemptId + " : "
+          + allocation + " with cluster resource : " + getClusterResource());
+    }
     return allocation;
   }
 
@@ -1442,10 +1444,17 @@ public class CapacityScheduler extends
   private CSAssignment allocateContainerOnSingleNode(
       CandidateNodeSet<FiCaSchedulerNode> candidates, FiCaSchedulerNode node,
       boolean withNodeHeartbeat) {
+    if (LOG.isDebugEnabled()) {
+      LOG.debug(
+          "Trying to schedule on node: " + node.getNodeName() + ", available: "
+              + node.getUnallocatedResource());
+    }
+
     // Backward compatible way to make sure previous behavior which allocation
     // driven by node heartbeat works.
     if (getNode(node.getNodeID()) != node) {
-      LOG.error("Trying to schedule on a removed node, please double check.");
+      LOG.error("Trying to schedule on a removed node, please double check, "
+          + "nodeId=" + node.getNodeID());
       return null;
     }
 
@@ -1459,14 +1468,19 @@ public class CapacityScheduler extends
       FiCaSchedulerApp reservedApplication = getCurrentAttemptForContainer(
           reservedContainer.getContainerId());
       if (reservedApplication == null) {
-        LOG.error("Trying to schedule for a finished app, please double check.");
+        LOG.error(
+            "Trying to schedule for a finished app, please double check. nodeId="
+                + node.getNodeID() + " container=" + reservedContainer
+                .getContainerId());
         return null;
       }
 
       // Try to fulfill the reservation
-      LOG.info(
-          "Trying to fulfill reservation for application " + reservedApplication
-              .getApplicationId() + " on node: " + node.getNodeID());
+      if (LOG.isDebugEnabled()) {
+        LOG.debug("Trying to fulfill reservation for application "
+            + reservedApplication.getApplicationId() + " on node: " + node
+            .getNodeID());
+      }
 
       LeafQueue queue = ((LeafQueue) reservedApplication.getQueue());
       assignment = queue.assignContainers(getClusterResource(), candidates,
@@ -1530,12 +1544,6 @@ public class CapacityScheduler extends
       return null;
     }
 
-    if (LOG.isDebugEnabled()) {
-      LOG.debug(
-          "Trying to schedule on node: " + node.getNodeName() + ", available: "
-              + node.getUnallocatedResource());
-    }
-
     return allocateOrReserveNewContainers(candidates, withNodeHeartbeat);
   }
 
@@ -2851,6 +2859,11 @@ public class CapacityScheduler extends
           LOG.info("Failed to accept allocation proposal");
         }
 
+        if (LOG.isDebugEnabled()) {
+          LOG.debug("Allocation proposal accepted=" + isSuccess + ", proposal="
+              + request);
+        }
+
         // Update unconfirmed allocated resource.
         if (updateUnconfirmedAllocatedResource) {
           app.decUnconfirmedRes(request.getTotalAllocatedResource());

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/ParentQueue.java

@@ -90,6 +90,8 @@ public class ParentQueue extends AbstractCSQueue {
 
   private QueueOrderingPolicy queueOrderingPolicy;
 
+  private long lastSkipQueueDebugLoggingTimestamp = -1;
+
   public ParentQueue(CapacitySchedulerContext cs,
       String queueName, CSQueue parent, CSQueue old) throws IOException {
     super(cs, queueName, parent, old);
@@ -539,9 +541,14 @@ public class ParentQueue extends AbstractCSQueue {
     if (schedulingMode == SchedulingMode.RESPECT_PARTITION_EXCLUSIVITY
         && !accessibleToPartition(candidates.getPartition())) {
       if (LOG.isDebugEnabled()) {
-        LOG.debug("Skip this queue=" + getQueuePath()
-            + ", because it is not able to access partition=" + candidates
-            .getPartition());
+        long now = System.currentTimeMillis();
+        // Do logging every 1 sec to avoid excessive logging.
+        if (now - this.lastSkipQueueDebugLoggingTimestamp > 1000) {
+          LOG.debug("Skip this queue=" + getQueuePath()
+              + ", because it is not able to access partition=" + candidates
+              .getPartition());
+          this.lastSkipQueueDebugLoggingTimestamp = now;
+        }
       }
 
       ActivitiesLogger.QUEUE.recordQueueActivity(activitiesManager, node,
@@ -561,10 +568,15 @@ public class ParentQueue extends AbstractCSQueue {
     if (!super.hasPendingResourceRequest(candidates.getPartition(),
         clusterResource, schedulingMode)) {
       if (LOG.isDebugEnabled()) {
-        LOG.debug("Skip this queue=" + getQueuePath()
-            + ", because it doesn't need more resource, schedulingMode="
-            + schedulingMode.name() + " node-partition=" + candidates
-            .getPartition());
+        long now = System.currentTimeMillis();
+        // Do logging every 1 sec to avoid excessive logging.
+        if (now - this.lastSkipQueueDebugLoggingTimestamp > 1000) {
+          LOG.debug("Skip this queue=" + getQueuePath()
+              + ", because it doesn't need more resource, schedulingMode="
+              + schedulingMode.name() + " node-partition=" + candidates
+              .getPartition());
+          this.lastSkipQueueDebugLoggingTimestamp = now;
+        }
       }
 
       ActivitiesLogger.QUEUE.recordQueueActivity(activitiesManager, node,
@@ -666,12 +678,12 @@ public class ParentQueue extends AbstractCSQueue {
         assignment.setIncreasedAllocation(
             assignedToChild.isIncreasedAllocation());
 
-        LOG.info("assignedContainer" + " queue=" + getQueueName()
-            + " usedCapacity=" + getUsedCapacity() + " absoluteUsedCapacity="
-            + getAbsoluteUsedCapacity() + " used=" + queueUsage.getUsed()
-            + " cluster=" + clusterResource);
-
         if (LOG.isDebugEnabled()) {
+          LOG.debug("assignedContainer reserved=" + isReserved + " queue="
+              + getQueueName() + " usedCapacity=" + getUsedCapacity()
+              + " absoluteUsedCapacity=" + getAbsoluteUsedCapacity() + " used="
+              + queueUsage.getUsed() + " cluster=" + clusterResource);
+
           LOG.debug(
               "ParentQ=" + getQueueName() + " assignedSoFarInThisIteration="
                   + assignment.getResource() + " usedCapacity="

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/allocator/AbstractContainerAllocator.java

@@ -93,11 +93,14 @@ public abstract class AbstractContainerAllocator {
       assignment.setType(result.getContainerNodeType());
 
       if (result.getAllocationState() == AllocationState.RESERVED) {
-        // This is a reserved container
-        LOG.info("Reserved container " + " application="
-            + application.getApplicationId() + " resource=" + allocatedResource
-            + " queue=" + appInfo.getQueueName()
-            + " cluster=" + clusterResource);
+        if (LOG.isDebugEnabled()) {
+          // This is a reserved container
+          // Since re-reservation could happen again and again for already
+          // reserved containers. only do this in debug log.
+          LOG.debug("Reserved container " + " application=" + application
+              .getApplicationId() + " resource=" + allocatedResource + " queue="
+              + appInfo.getQueueName() + " cluster=" + clusterResource);
+        }
         assignment.getAssignmentInformation().addReservationDetails(
             updatedContainer, application.getCSLeafQueue().getQueuePath());
         assignment.getAssignmentInformation().incrReservations();

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/common/fica/FiCaSchedulerApp.java

@@ -607,6 +607,11 @@ public class FiCaSchedulerApp extends SchedulerApplicationAttempt {
               schedulerContainer.getRmContainer(),
               schedulerContainer.getRmContainer().getContainer(),
               reReservation);
+
+          LOG.info("Reserved container=" + rmContainer.getContainerId()
+              + ", on node=" + schedulerContainer.getSchedulerNode()
+              + " with resource=" + rmContainer
+              .getAllocatedOrReservedResource());
         }
       }
     } finally {