From c3c86b18cb40b0e8437016b1418d2dbe29c0d8c3 Mon Sep 17 00:00:00 2001 From: Jim Brennan Date: Mon, 28 Jun 2021 14:52:19 +0000 Subject: [PATCH] YARN-10834. Intra-queue preemption: apps that don't use defined custom resource won't be preempted. Contributed by Eric Payne. (cherry picked from commit dc6f456e953e685370277d3d6bf3515b5001bca3) --- .../FifoIntraQueuePreemptionPlugin.java | 5 +- ...cityPreemptionPolicyIntraQueueWithDRF.java | 83 +++++++++++++++++++ 2 files changed, 86 insertions(+), 2 deletions(-) diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/monitor/capacity/FifoIntraQueuePreemptionPlugin.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/monitor/capacity/FifoIntraQueuePreemptionPlugin.java index 2e69b9f1b15..6e6d662fa14 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/monitor/capacity/FifoIntraQueuePreemptionPlugin.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/monitor/capacity/FifoIntraQueuePreemptionPlugin.java @@ -290,8 +290,9 @@ public class FifoIntraQueuePreemptionPlugin // Once unallocated resource is 0, we can stop assigning ideal per app. if (Resources.lessThanOrEqual(rc, clusterResource, - queueReassignableResource, Resources.none()) || rc - .isAnyMajorResourceZeroOrNegative(queueReassignableResource)) { + queueReassignableResource, Resources.none()) || + (rc.isAnyMajorResourceZeroOrNegative(queueReassignableResource) + && context.getInQueuePreemptionConservativeDRF())) { continue; } diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/monitor/capacity/TestProportionalCapacityPreemptionPolicyIntraQueueWithDRF.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/monitor/capacity/TestProportionalCapacityPreemptionPolicyIntraQueueWithDRF.java index cfc15d4cb7c..dbaae1885b5 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/monitor/capacity/TestProportionalCapacityPreemptionPolicyIntraQueueWithDRF.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/monitor/capacity/TestProportionalCapacityPreemptionPolicyIntraQueueWithDRF.java @@ -18,11 +18,14 @@ package org.apache.hadoop.yarn.server.resourcemanager.monitor.capacity; +import org.apache.hadoop.yarn.api.protocolrecords.ResourceTypes; import org.apache.hadoop.yarn.api.records.Resource; +import org.apache.hadoop.yarn.api.records.ResourceInformation; import org.apache.hadoop.yarn.server.resourcemanager.monitor.capacity.mockframework.ProportionalCapacityPreemptionPolicyMockFramework; import org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.CapacitySchedulerConfiguration; import org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.LeafQueue; import org.apache.hadoop.yarn.util.resource.DominantResourceCalculator; +import org.apache.hadoop.yarn.util.resource.ResourceUtils; import org.junit.Before; import org.junit.Test; @@ -189,6 +192,86 @@ public class TestProportionalCapacityPreemptionPolicyIntraQueueWithDRF getAppAttemptId(1)))); } + @SuppressWarnings("unchecked") + @Test + public void testIntraQueuePreemptionFairOrdering3ResourcesWithStrictAndRelaxedDRF() + throws IOException { + /** + * Continue to allow intra-queue preemption when only one of the user's + * resources is above the user limit. + * Queue structure is: + * + * 
+     *       root
+     *     /  |
+     *    a   b
+     *
+ * + * Guaranteed resource of a and b are 30720:300 and 30720:300 Total cluster + * resource = 61440:600. + * Scenario: Queue B has one running app using 61720:60 resources with no + * pending resources, and one app with no used resources and 30720:30 + * pending resources. + * + * The first part of the test is to show what happens when the conservative + * DRF property is set. Since the memory is above and the vcores is below + * the user limit, only the minimum number of containers is allowed. + * In the second part, since conservative DRF is relaxed, all containers + * needed are allowed to be preempted (minus the AM size). + */ + + conf.set(CapacitySchedulerConfiguration.INTRAQUEUE_PREEMPTION_ORDER_POLICY, + "userlimit_first"); + conf.set(CapacitySchedulerConfiguration.PREFIX + + "root.b." + CapacitySchedulerConfiguration.ORDERING_POLICY, "fair"); + conf.setBoolean( + CapacitySchedulerConfiguration.IN_QUEUE_PREEMPTION_CONSERVATIVE_DRF, + true); + String RESOURCE_1 = "res1"; + riMap.put(RESOURCE_1, ResourceInformation + .newInstance(RESOURCE_1, "", 0, ResourceTypes.COUNTABLE, 0, + Integer.MAX_VALUE)); + + ResourceUtils.initializeResourcesFromResourceInformationMap(riMap); + + String labelsConfig = "=61440:600,true;"; + String nodesConfig = // n1 has no label + "n1= res=61440:600"; + String queuesConfig = + // guaranteed,max,used,pending,reserved + "root(=[61440:600 61440:600 61440:600 30720:30 0]);" + // root + "-a(=[30720:300 61440:600 0:0 0:0 0]);" + // a + "-b(=[30720:300 61440:600 61440:60 30720:30 0]);"; // b + + String appsConfig = + "b\t" + "(1,1024:1,n1,,60,false,0:0,user1);" + // app1 in b + "b\t" + "(1,0:0,n1,,0,false,30720:30,user3);"; // app2 in b + + buildEnv(labelsConfig, nodesConfig, queuesConfig, appsConfig); + Resource ul = Resource.newInstance(30720, 300); + when(((LeafQueue)(cs.getQueue("root.b"))) + .getResourceLimitForAllUsers(any(), any(), any(), any()) + ).thenReturn(ul); + policy.editSchedule(); + + verify(eventHandler, times(0)).handle(argThat( + new TestProportionalCapacityPreemptionPolicy.IsPreemptionRequestFor( + getAppAttemptId(1)))); + reset(eventHandler); + + conf.setBoolean( + CapacitySchedulerConfiguration.IN_QUEUE_PREEMPTION_CONSERVATIVE_DRF, + false); + buildEnv(labelsConfig, nodesConfig, queuesConfig, appsConfig); + when(((LeafQueue)(cs.getQueue("root.b"))) + .getResourceLimitForAllUsers(any(), any(), any(), any()) + ).thenReturn(ul); + policy.editSchedule(); + verify(eventHandler, times(29)).handle(argThat( + new TestProportionalCapacityPreemptionPolicy.IsPreemptionRequestFor( + getAppAttemptId(1)))); + } + @Test public void testIntraQueuePreemptionWithDominantVCoreResource() throws IOException {