Matthias Andreas Benkard | 832a54e | 2019-01-29 09:27:38 +0100 | [diff] [blame] | 1 | // Copyright 2018 The Kubernetes Authors. |
| 2 | // |
| 3 | // Licensed under the Apache License, Version 2.0 (the "License"); |
| 4 | // you may not use this file except in compliance with the License. |
| 5 | // You may obtain a copy of the License at |
| 6 | // |
| 7 | // http://www.apache.org/licenses/LICENSE-2.0 |
| 8 | // |
| 9 | // Unless required by applicable law or agreed to in writing, software |
| 10 | // distributed under the License is distributed on an "AS IS" BASIS, |
| 11 | // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 12 | // See the License for the specific language governing permissions and |
| 13 | // limitations under the License. |
| 14 | |
| 15 | package manager |
| 16 | |
| 17 | import ( |
| 18 | "context" |
| 19 | "fmt" |
| 20 | "net/http" |
| 21 | "sync" |
| 22 | "time" |
| 23 | |
| 24 | utilmetrics "github.com/kubernetes-incubator/metrics-server/pkg/metrics" |
| 25 | "github.com/kubernetes-incubator/metrics-server/pkg/sink" |
| 26 | "github.com/kubernetes-incubator/metrics-server/pkg/sources" |
| 27 | |
| 28 | "github.com/golang/glog" |
| 29 | "github.com/prometheus/client_golang/prometheus" |
| 30 | ) |
| 31 | |
| 32 | var ( |
| 33 | // initialized below to an actual value by a call to RegisterTickDuration |
| 34 | // (acts as a no-op by default), but we can't just register it in the constructor, |
| 35 | // since it could be called multiple times during setup. |
| 36 | tickDuration prometheus.Histogram = prometheus.NewHistogram(prometheus.HistogramOpts{}) |
| 37 | ) |
| 38 | |
| 39 | // RegisterTickDuration creates and registers a histogram metric for |
| 40 | // scrape duration, suitable for use in the overall manager. |
| 41 | func RegisterDurationMetrics(resolution time.Duration) { |
| 42 | tickDuration = prometheus.NewHistogram( |
| 43 | prometheus.HistogramOpts{ |
| 44 | Namespace: "metrics_server", |
| 45 | Subsystem: "manager", |
| 46 | Name: "tick_duration_seconds", |
| 47 | Help: "The total time spent collecting and storing metrics in seconds.", |
| 48 | Buckets: utilmetrics.BucketsForScrapeDuration(resolution), |
| 49 | }, |
| 50 | ) |
| 51 | prometheus.MustRegister(tickDuration) |
| 52 | } |
| 53 | |
| 54 | // Runnable represents something that can be run until a signal is given to stop. |
| 55 | type Runnable interface { |
| 56 | // Run runs this runnable until the given channel is closed. |
| 57 | // It should not block -- it will spawn its own goroutine. |
| 58 | RunUntil(stopCh <-chan struct{}) |
| 59 | } |
| 60 | |
| 61 | type Manager struct { |
| 62 | source sources.MetricSource |
| 63 | sink sink.MetricSink |
| 64 | resolution time.Duration |
| 65 | |
| 66 | healthMu sync.RWMutex |
| 67 | lastTickStart time.Time |
| 68 | lastOk bool |
| 69 | } |
| 70 | |
| 71 | func NewManager(metricSrc sources.MetricSource, metricSink sink.MetricSink, resolution time.Duration) *Manager { |
| 72 | manager := Manager{ |
| 73 | source: metricSrc, |
| 74 | sink: metricSink, |
| 75 | resolution: resolution, |
| 76 | } |
| 77 | |
| 78 | return &manager |
| 79 | } |
| 80 | |
| 81 | func (rm *Manager) RunUntil(stopCh <-chan struct{}) { |
| 82 | go func() { |
| 83 | ticker := time.NewTicker(rm.resolution) |
| 84 | defer ticker.Stop() |
| 85 | rm.Collect(time.Now()) |
| 86 | |
| 87 | for { |
| 88 | select { |
| 89 | case startTime := <-ticker.C: |
| 90 | rm.Collect(startTime) |
| 91 | case <-stopCh: |
| 92 | return |
| 93 | } |
| 94 | } |
| 95 | }() |
| 96 | } |
| 97 | |
| 98 | func (rm *Manager) Collect(startTime time.Time) { |
| 99 | rm.healthMu.Lock() |
| 100 | rm.lastTickStart = startTime |
| 101 | rm.healthMu.Unlock() |
| 102 | |
| 103 | healthyTick := true |
| 104 | |
| 105 | ctx, cancelTimeout := context.WithTimeout(context.Background(), rm.resolution) |
| 106 | defer cancelTimeout() |
| 107 | |
| 108 | glog.V(6).Infof("Beginning cycle, collecting metrics...") |
| 109 | data, collectErr := rm.source.Collect(ctx) |
| 110 | if collectErr != nil { |
| 111 | glog.Errorf("unable to fully collect metrics: %v", collectErr) |
| 112 | |
| 113 | // only consider this an indication of bad health if we |
| 114 | // couldn't collect from any nodes -- one node going down |
| 115 | // shouldn't indicate that metrics-server is unhealthy |
| 116 | if len(data.Nodes) == 0 { |
| 117 | healthyTick = false |
| 118 | } |
| 119 | |
| 120 | // NB: continue on so that we don't lose all metrics |
| 121 | // if one node goes down |
| 122 | } |
| 123 | |
| 124 | glog.V(6).Infof("...Storing metrics...") |
| 125 | recvErr := rm.sink.Receive(data) |
| 126 | if recvErr != nil { |
| 127 | glog.Errorf("unable to save metrics: %v", recvErr) |
| 128 | |
| 129 | // any failure to save means we're unhealthy |
| 130 | healthyTick = false |
| 131 | } |
| 132 | |
| 133 | collectTime := time.Now().Sub(startTime) |
| 134 | tickDuration.Observe(float64(collectTime) / float64(time.Second)) |
| 135 | glog.V(6).Infof("...Cycle complete") |
| 136 | |
| 137 | rm.healthMu.Lock() |
| 138 | rm.lastOk = healthyTick |
| 139 | rm.healthMu.Unlock() |
| 140 | } |
| 141 | |
| 142 | // CheckHealth checks the health of the manager by looking at tick times, |
| 143 | // and checking if we have at least one node in the collected data. |
| 144 | // It implements the health checker func part of the healthz checker. |
| 145 | func (rm *Manager) CheckHealth(_ *http.Request) error { |
| 146 | rm.healthMu.RLock() |
| 147 | lastTick := rm.lastTickStart |
| 148 | healthyTick := rm.lastOk |
| 149 | rm.healthMu.RUnlock() |
| 150 | |
| 151 | // use 1.1 for a bit of wiggle room |
| 152 | maxTickWait := time.Duration(1.1 * float64(rm.resolution)) |
| 153 | tickWait := time.Now().Sub(lastTick) |
| 154 | |
| 155 | if tickWait > maxTickWait { |
| 156 | return fmt.Errorf("time since last tick (%s) was greater than expected metrics resolution (%s)", tickWait, maxTickWait) |
| 157 | } |
| 158 | |
| 159 | if !healthyTick { |
| 160 | return fmt.Errorf("there was an error collecting or saving metrics in the last collection tick") |
| 161 | } |
| 162 | |
| 163 | return nil |
| 164 | } |