com.netflix.genie.web.services.loadbalancers.script

Class ScriptLoadBalancer

java.lang.Object

com.netflix.genie.web.services.loadbalancers.script.ScriptLoadBalancer

All Implemented Interfaces:

ClusterLoadBalancer

public class ScriptLoadBalancer
extends java.lang.Object
implements ClusterLoadBalancer

An implementation of the ClusterLoadBalancer interface which uses user a supplied script to make decisions based on the list of clusters and the job request supplied.

The contract between the script and the Java code is that the script will be supplied global variables clusters and jobRequest which will be JSON strings representing the list (array) of clusters matching the cluster criteria tags and the job request that kicked off this evaluation. The code expects the script to either return the id of the cluster if one is selected or null if none was selected.

Note: this LoadBalancer implementation intentionally returns 'null' (a.k.a. 'no preference') in case of error, rather throwing an exception. For example if the script cannot be loaded, or if an invalid cluster is returned. TODO: this logic of falling back to 'no preference' in case of error should be moved out of this implementation and into the service using this interface.

Since:

3.1.0

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
protected static class	ScriptLoadBalancer.Evaluator
protected static class	ScriptLoadBalancer.Loader

Constructor Summary

Constructors

Constructor and Description
ScriptLoadBalancer(org.springframework.core.task.AsyncTaskExecutor asyncTaskExecutor, org.springframework.scheduling.TaskScheduler taskScheduler, GenieFileTransferService fileTransferService, org.springframework.core.env.Environment environment, com.fasterxml.jackson.databind.ObjectMapper mapper, io.micrometer.core.instrument.MeterRegistry registry) Constructor.

Method Summary

Modifier and Type	Method and Description
Cluster	selectCluster(@NonNull @NotEmpty java.util.Set<Cluster> clusters, @NonNull JobRequest jobRequest) Return best cluster to run job on.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

ScriptLoadBalancer

public ScriptLoadBalancer(org.springframework.core.task.AsyncTaskExecutor asyncTaskExecutor,
                          org.springframework.scheduling.TaskScheduler taskScheduler,
                          GenieFileTransferService fileTransferService,
                          org.springframework.core.env.Environment environment,
                          com.fasterxml.jackson.databind.ObjectMapper mapper,
                          io.micrometer.core.instrument.MeterRegistry registry)

Constructor.

Parameters:

asyncTaskExecutor - The asynchronous task executor to use to run the load balancer script in

taskScheduler - The task scheduler to schedule the script refresh task with

fileTransferService - The file transfer service to use to download the script

environment - The program environment to get properties from

mapper - The object mapper to use to serialize objects to JSON for binding with scripts

registry - The metrics registry to use for collecting metrics

Method Detail

selectCluster

public Cluster selectCluster(@Nonnull @NonNull @NotEmpty
                             @NonNull @NotEmpty java.util.Set<Cluster> clusters,
                             @Nonnull @NonNull
                             @NonNull JobRequest jobRequest)
                      throws GenieException

Return best cluster to run job on.

Specified by:

selectCluster in interface ClusterLoadBalancer

Parameters:

clusters - An immutable, non-empty list of available clusters to choose from

jobRequest - The job request these clusters are being load balanced for

Returns:

the "best" cluster to run job on or null if no cluster selected

Throws:

GenieException - if there is any error