1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
|
// Copyright Vespa.ai. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
package com.yahoo.vespa.hosted.provision.maintenance;
import com.yahoo.config.provision.ApplicationId;
import com.yahoo.config.provision.ApplicationLockException;
import com.yahoo.config.provision.ClusterResources;
import com.yahoo.config.provision.ClusterSpec;
import com.yahoo.jdisc.Metric;
import com.yahoo.transaction.Mutex;
import com.yahoo.vespa.hosted.provision.NodeList;
import com.yahoo.vespa.hosted.provision.NodeRepository;
import com.yahoo.vespa.hosted.provision.applications.Application;
import com.yahoo.vespa.hosted.provision.applications.Applications;
import com.yahoo.vespa.hosted.provision.applications.Cluster;
import com.yahoo.vespa.hosted.provision.autoscale.Autoscaler;
import com.yahoo.vespa.hosted.provision.autoscale.Autoscaling;
import java.time.Duration;
import java.util.List;
import java.util.Map;
import java.util.Optional;
/**
* Maintainer computing scaling suggestions for all clusters
*
* @author bratseth
*/
public class ScalingSuggestionsMaintainer extends NodeRepositoryMaintainer {
private final Autoscaler autoscaler;
public ScalingSuggestionsMaintainer(NodeRepository nodeRepository,
Duration interval,
Metric metric) {
super(nodeRepository, interval, metric);
this.autoscaler = new Autoscaler(nodeRepository);
}
@Override
protected double maintain() {
if ( ! nodeRepository().zone().environment().isProduction()) return 1.0;
int attempts = 0;
int failures = 0;
outer:
for (var application : activeNodesByApplication().entrySet()) {
for (var cluster : nodesByCluster(application.getValue()).entrySet()) {
if (shuttingDown()) break outer;
attempts++;
if ( ! suggest(application.getKey(), cluster.getKey(), cluster.getValue()))
failures++;
}
}
return asSuccessFactorDeviation(attempts, failures);
}
private Applications applications() {
return nodeRepository().applications();
}
private boolean suggest(ApplicationId applicationId,
ClusterSpec.Id clusterId,
NodeList clusterNodes) {
Application application = applications().get(applicationId).orElse(Application.empty(applicationId));
Optional<Cluster> cluster = application.cluster(clusterId);
if (cluster.isEmpty()) return true;
var suggestions = autoscaler.suggest(application, cluster.get(), clusterNodes);
if ( ! shouldUpdateSuggestion(cluster.get().suggestions(), suggestions))
return true;
// Wait only a short time for the lock to avoid interfering with change deployments
try (Mutex lock = nodeRepository().applications().lock(applicationId, Duration.ofSeconds(1))) {
applications().get(applicationId).ifPresent(a -> updateSuggestion(suggestions, clusterId, a, lock));
return true;
}
catch (ApplicationLockException e) {
return false;
}
}
private boolean shouldUpdateSuggestion(List<Autoscaling> currentSuggestions, List<Autoscaling> newSuggestions) {
// Only compare previous best suggestion with current best suggestion
var currentSuggestion = currentSuggestions.stream().findFirst();
var newSuggestion = newSuggestions.stream().findFirst();
if (currentSuggestion.isEmpty()) return true;
if (newSuggestion.isEmpty()) return false;
return newSuggestion.get().status() != Autoscaling.Status.waiting
&& (currentSuggestion.get().resources().isEmpty()
|| currentSuggestion.get().at().isBefore(nodeRepository().clock().instant().minus(Duration.ofDays(7)))
|| (newSuggestion.get().resources().isPresent() && isHigher(newSuggestion.get().resources().get(), currentSuggestion.get().resources().get())));
}
private void updateSuggestion(List<Autoscaling> suggestions,
ClusterSpec.Id clusterId,
Application application,
Mutex lock) {
Optional<Cluster> cluster = application.cluster(clusterId);
if (cluster.isEmpty()) return;
applications().put(application.with(cluster.get().withSuggestions(suggestions)), lock);
}
private boolean isHigher(ClusterResources r1, ClusterResources r2) {
// Use cost as a measure of "highness" over multiple dimensions
return r1.totalResources().cost() > r2.totalResources().cost();
}
private Map<ClusterSpec.Id, NodeList> nodesByCluster(NodeList applicationNodes) {
return applicationNodes.groupingBy(n -> n.allocation().get().membership().cluster().id());
}
}
|
