com.netflix.fenzo.OptimizingShortfallEvaluator Maven / Gradle / Ivy
/*
* Copyright 2017 Netflix, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package com.netflix.fenzo;
import com.netflix.fenzo.queues.InternalTaskQueue;
import com.netflix.fenzo.queues.InternalTaskQueues;
import com.netflix.fenzo.queues.QueuableTask;
import java.util.Collections;
import java.util.List;
import java.util.Map;
import java.util.Set;
/**
* This class implements an optimal evaluation of the shortfall of the right VMs needed for cluster scale up.
* In a cluster with multiple VM groups, this evaluator is able to determine the minimum number of VMs of each group
* required to satisfy the resource demands from pending tasks at the end of a scheduling iteration.
*
* This evaluator can only be used with {@link TaskSchedulingService}, it cannot be used when directly using
* {@link TaskScheduler}. The latter uses {@link NaiveShortfallEvaluator} for shortfall analysis.
*
* This evaluator works by requesting a "pseudo" scheduling iteration from {@link TaskSchedulingService} with a
* new queue that is cloned from the original queue being used and adding to it only the tasks that failed assignments.
* {@link TaskSchedulingService} performs the pseudo scheduling run by creating the appropriate number of pseudo VMs
* for each group of VMs. For this, the autoscale rules are consulted to ensure the scale up will honor the maximum VMs
* for each group.
*
* The pseudo scheduling run performs an entire scheduling iteration using the cloned queue and pseudo VMs in addition
* to any new VM leases that have been added since previous scheduling iteration. This will invoke any and all task
* constraints as well as fitness function setup in the scheduler. The scheduling result is used to determine the
* number of VMs in each group and then the results are discarded. As expected, the pseudo scheduling run has no impact
* on the real scheduling assignments made.
*
* Tasks for which scale up is requested by this evaluator are remembered and not requested again until certain delay.
*/
class OptimizingShortfallEvaluator extends BaseShortfallEvaluator {
@Override
public Map getShortfall(Set vmGroupNames, Set failures, AutoScaleRules autoScaleRules) {
if (schedulingService == null || failures == null || failures.isEmpty())
return Collections.emptyMap();
final List filteredTasks = filterFailedTasks(failures);
final Map shortfallTasksPerGroup = fillShortfallMap(vmGroupNames, filteredTasks);
if (shortfallTasksPerGroup.isEmpty())
return Collections.emptyMap();
if (schedulingService.isShutdown())
return Collections.emptyMap();
final InternalTaskQueue taskQueue = createAndFillAlternateQueue(filteredTasks);
return schedulingService.requestPseudoScheduling(taskQueue, shortfallTasksPerGroup);
}
private InternalTaskQueue createAndFillAlternateQueue(List shortfallTasks) {
final InternalTaskQueue taskQueue = InternalTaskQueues.createQueueOf(schedulingService.getQueue());
for (TaskRequest t: shortfallTasks) {
taskQueue.queueTask((QueuableTask) t);
}
return taskQueue;
}
}