|
| 1 | +/* |
| 2 | +Copyright 2017 The Kubernetes Authors. |
| 3 | +
|
| 4 | +Licensed under the Apache License, Version 2.0 (the "License"); |
| 5 | +you may not use this file except in compliance with the License. |
| 6 | +You may obtain a copy of the License at |
| 7 | +
|
| 8 | + http://www.apache.org/licenses/LICENSE-2.0 |
| 9 | +
|
| 10 | +Unless required by applicable law or agreed to in writing, software |
| 11 | +distributed under the License is distributed on an "AS IS" BASIS, |
| 12 | +WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 13 | +See the License for the specific language governing permissions and |
| 14 | +limitations under the License. |
| 15 | +*/ |
| 16 | + |
| 17 | +package app |
| 18 | + |
| 19 | +import ( |
| 20 | + "context" |
| 21 | + "fmt" |
| 22 | + "net/http" |
| 23 | + "os" |
| 24 | + "time" |
| 25 | + |
| 26 | + "github.com/golang/glog" |
| 27 | + "github.com/prometheus/client_golang/prometheus/promhttp" |
| 28 | + "volcano.sh/volcano/cmd/kube-batch/app/options" |
| 29 | + "volcano.sh/volcano/pkg/scheduler" |
| 30 | + "volcano.sh/volcano/pkg/version" |
| 31 | + |
| 32 | + v1 "k8s.io/api/core/v1" |
| 33 | + "k8s.io/apimachinery/pkg/util/uuid" |
| 34 | + clientset "k8s.io/client-go/kubernetes" |
| 35 | + "k8s.io/client-go/kubernetes/scheme" |
| 36 | + corev1 "k8s.io/client-go/kubernetes/typed/core/v1" |
| 37 | + |
| 38 | + // Register gcp auth |
| 39 | + _ "k8s.io/client-go/plugin/pkg/client/auth/gcp" |
| 40 | + "k8s.io/client-go/rest" |
| 41 | + restclient "k8s.io/client-go/rest" |
| 42 | + "k8s.io/client-go/tools/clientcmd" |
| 43 | + "k8s.io/client-go/tools/leaderelection" |
| 44 | + "k8s.io/client-go/tools/leaderelection/resourcelock" |
| 45 | + "k8s.io/client-go/tools/record" |
| 46 | +) |
| 47 | + |
| 48 | +const ( |
| 49 | + leaseDuration = 15 * time.Second |
| 50 | + renewDeadline = 10 * time.Second |
| 51 | + retryPeriod = 5 * time.Second |
| 52 | + apiVersion = "v1alpha1" |
| 53 | +) |
| 54 | + |
| 55 | +func buildConfig(opt *options.ServerOption) (*rest.Config, error) { |
| 56 | + var cfg *rest.Config |
| 57 | + var err error |
| 58 | + |
| 59 | + master := opt.Master |
| 60 | + kubeconfig := opt.Kubeconfig |
| 61 | + if master != "" || kubeconfig != "" { |
| 62 | + cfg, err = clientcmd.BuildConfigFromFlags(master, kubeconfig) |
| 63 | + } else { |
| 64 | + cfg, err = rest.InClusterConfig() |
| 65 | + } |
| 66 | + if err != nil { |
| 67 | + return nil, err |
| 68 | + } |
| 69 | + cfg.QPS = opt.KubeAPIQPS |
| 70 | + cfg.Burst = opt.KubeAPIBurst |
| 71 | + |
| 72 | + return cfg, nil |
| 73 | +} |
| 74 | + |
| 75 | +// Run the kubeBatch scheduler |
| 76 | +func Run(opt *options.ServerOption) error { |
| 77 | + if opt.PrintVersion { |
| 78 | + version.PrintVersionAndExit() |
| 79 | + } |
| 80 | + |
| 81 | + config, err := buildConfig(opt) |
| 82 | + if err != nil { |
| 83 | + return err |
| 84 | + } |
| 85 | + |
| 86 | + // Start policy controller to allocate resources. |
| 87 | + sched, err := scheduler.NewScheduler(config, |
| 88 | + opt.SchedulerName, |
| 89 | + opt.SchedulerConf, |
| 90 | + opt.SchedulePeriod, |
| 91 | + opt.DefaultQueue) |
| 92 | + if err != nil { |
| 93 | + panic(err) |
| 94 | + } |
| 95 | + |
| 96 | + go func() { |
| 97 | + http.Handle("/metrics", promhttp.Handler()) |
| 98 | + glog.Fatalf("Prometheus Http Server failed %s", http.ListenAndServe(opt.ListenAddress, nil)) |
| 99 | + }() |
| 100 | + |
| 101 | + run := func(ctx context.Context) { |
| 102 | + sched.Run(ctx.Done()) |
| 103 | + <-ctx.Done() |
| 104 | + } |
| 105 | + |
| 106 | + if !opt.EnableLeaderElection { |
| 107 | + run(context.TODO()) |
| 108 | + return fmt.Errorf("finished without leader elect") |
| 109 | + } |
| 110 | + |
| 111 | + leaderElectionClient, err := clientset.NewForConfig(restclient.AddUserAgent(config, "leader-election")) |
| 112 | + if err != nil { |
| 113 | + return err |
| 114 | + } |
| 115 | + |
| 116 | + // Prepare event clients. |
| 117 | + broadcaster := record.NewBroadcaster() |
| 118 | + broadcaster.StartRecordingToSink(&corev1.EventSinkImpl{Interface: leaderElectionClient.CoreV1().Events(opt.LockObjectNamespace)}) |
| 119 | + eventRecorder := broadcaster.NewRecorder(scheme.Scheme, v1.EventSource{Component: opt.SchedulerName}) |
| 120 | + |
| 121 | + hostname, err := os.Hostname() |
| 122 | + if err != nil { |
| 123 | + return fmt.Errorf("unable to get hostname: %v", err) |
| 124 | + } |
| 125 | + // add a uniquifier so that two processes on the same host don't accidentally both become active |
| 126 | + id := hostname + "_" + string(uuid.NewUUID()) |
| 127 | + |
| 128 | + rl, err := resourcelock.New(resourcelock.ConfigMapsResourceLock, |
| 129 | + opt.LockObjectNamespace, |
| 130 | + "kube-batch", |
| 131 | + leaderElectionClient.CoreV1(), |
| 132 | + resourcelock.ResourceLockConfig{ |
| 133 | + Identity: id, |
| 134 | + EventRecorder: eventRecorder, |
| 135 | + }) |
| 136 | + if err != nil { |
| 137 | + return fmt.Errorf("couldn't create resource lock: %v", err) |
| 138 | + } |
| 139 | + |
| 140 | + leaderelection.RunOrDie(context.TODO(), leaderelection.LeaderElectionConfig{ |
| 141 | + Lock: rl, |
| 142 | + LeaseDuration: leaseDuration, |
| 143 | + RenewDeadline: renewDeadline, |
| 144 | + RetryPeriod: retryPeriod, |
| 145 | + Callbacks: leaderelection.LeaderCallbacks{ |
| 146 | + OnStartedLeading: run, |
| 147 | + OnStoppedLeading: func() { |
| 148 | + glog.Fatalf("leaderelection lost") |
| 149 | + }, |
| 150 | + }, |
| 151 | + }) |
| 152 | + return fmt.Errorf("lost lease") |
| 153 | +} |
0 commit comments