diff --git hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSLeafQueue.java hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSLeafQueue.java index aad291619c..7c64435562 100644 --- hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSLeafQueue.java +++ hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSLeafQueue.java @@ -198,18 +198,20 @@ public void collectSchedulerApplications( } @Override - public void updateInternal(boolean checkStarvation) { + public void updateInternal() { readLock.lock(); try { policy.computeShares(runnableApps, getFairShare()); - if (checkStarvation) { - updateStarvedApps(); - } } finally { readLock.unlock(); } } + @Override + void updateStarvationStats() { + updateStarvedApps(); + } + /** * Compute the extent of fairshare starvation for a set of apps. * @@ -283,6 +285,8 @@ private void updateStarvedAppsMinshare( * If this queue is starving due to fairshare, there must be at least * one application that is starved. And, even if the queue is not * starved due to fairshare, there might still be starved applications. + * + * Caller does not need read/write lock. */ private void updateStarvedApps() { // Fetch apps with pending demand diff --git hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSParentQueue.java hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSParentQueue.java index 1c8e9ced59..a701f49548 100644 --- hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSParentQueue.java +++ hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSParentQueue.java @@ -79,13 +79,25 @@ void removeChildQueue(FSQueue child) { } @Override - public void updateInternal(boolean checkStarvation) { + public void updateInternal() { readLock.lock(); try { policy.computeShares(childQueues, getFairShare()); for (FSQueue childQueue : childQueues) { childQueue.getMetrics().setFairShare(childQueue.getFairShare()); - childQueue.updateInternal(checkStarvation); + childQueue.updateInternal(); + } + } finally { + readLock.unlock(); + } + } + + @Override + void updateStarvationStats() { + readLock.lock(); + try { + for (FSQueue childQueue : childQueues) { + childQueue.updateStarvationStats(); } } finally { readLock.unlock(); diff --git hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSQueue.java hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSQueue.java index b5592c55bc..982b71ef68 100644 --- hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSQueue.java +++ hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSQueue.java @@ -323,16 +323,32 @@ public boolean isPreemptable() { /** * Recomputes the shares for all child queues and applications based on this - * queue's current share, and checks for starvation. + * queue's current share. * - * @param checkStarvation whether to check for fairshare or minshare - * starvation on update + * To be called holding the scheduler writelock. */ - abstract void updateInternal(boolean checkStarvation); + abstract void updateInternal(); - public void update(Resource fairShare, boolean checkStarvation) { + /** + * Update starvation stats for all child queues/apps and identify starved + * applications. + * + * To be called holding the scheduler readlock. + */ + abstract void updateStarvationStats(); + + + /** + * Set the queue's fairshare and update the demand/fairshare of child + * queues/applications. + * + * To be called holding the scheduler writelock. + * + * @param fairShare + */ + public void update(Resource fairShare) { setFairShare(fairShare); - updateInternal(checkStarvation); + updateInternal(); } /** diff --git hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FairScheduler.java hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FairScheduler.java index 4f3e4f9140..4b448afce6 100644 --- hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FairScheduler.java +++ hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FairScheduler.java @@ -351,34 +351,42 @@ public void run() { */ @VisibleForTesting public void update() { - try { - writeLock.lock(); - - FSQueue rootQueue = queueMgr.getRootQueue(); + FSQueue rootQueue = queueMgr.getRootQueue(); + // Update demands and fairshares + writeLock.lock(); + try { // Recursively update demands for all queues rootQueue.updateDemand(); - - Resource clusterResource = getClusterResource(); - rootQueue.update(clusterResource, shouldAttemptPreemption()); + rootQueue.update(getClusterResource()); // Update metrics updateRootQueueMetrics(); - - if (LOG.isDebugEnabled()) { - if (--updatesToSkipForDebug < 0) { - updatesToSkipForDebug = UPDATE_DEBUG_FREQUENCY; - LOG.debug("Cluster Capacity: " + clusterResource + - " Allocations: " + rootMetrics.getAllocatedResources() + - " Availability: " + Resource.newInstance( - rootMetrics.getAvailableMB(), - rootMetrics.getAvailableVirtualCores()) + - " Demand: " + rootQueue.getDemand()); - } - } } finally { writeLock.unlock(); } + + // Update starvation stats and identify starved applications + if (shouldAttemptPreemption()) { + readLock.lock(); + try { + rootQueue.updateStarvationStats(); + } finally { + readLock.unlock(); + } + } + + if (LOG.isDebugEnabled()) { + if (--updatesToSkipForDebug < 0) { + updatesToSkipForDebug = UPDATE_DEBUG_FREQUENCY; + LOG.debug("Cluster Capacity: " + getClusterResource() + + " Allocations: " + rootMetrics.getAllocatedResources() + + " Availability: " + Resource.newInstance( + rootMetrics.getAvailableMB(), + rootMetrics.getAvailableVirtualCores()) + + " Demand: " + rootQueue.getDemand()); + } + } } public RMContainerTokenSecretManager