|
| 1 | +// Copyright The OpenTelemetry 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 v1beta1 |
| 16 | + |
| 17 | +import ( |
| 18 | + "context" |
| 19 | + "fmt" |
| 20 | + "strings" |
| 21 | + |
| 22 | + "go.opentelemetry.io/otel/attribute" |
| 23 | + "go.opentelemetry.io/otel/exporters/prometheus" |
| 24 | + "go.opentelemetry.io/otel/metric" |
| 25 | + sdkmetric "go.opentelemetry.io/otel/sdk/metric" |
| 26 | + "sigs.k8s.io/controller-runtime/pkg/client" |
| 27 | + "sigs.k8s.io/controller-runtime/pkg/metrics" |
| 28 | +) |
| 29 | + |
| 30 | +const ( |
| 31 | + meterName = "crd-metrics" |
| 32 | +) |
| 33 | + |
| 34 | +// Metric labels |
| 35 | + |
| 36 | +const ( |
| 37 | + prefix = "opentelemetry_collector_" |
| 38 | + receiversMetricName = prefix + "receivers" |
| 39 | + exportersMetricName = prefix + "exporters" |
| 40 | + processorsMetricName = prefix + "processors" |
| 41 | + extensionsMetricName = prefix + "extensions" |
| 42 | + connectorsMetricName = prefix + "connectors" |
| 43 | + modeMetricName = prefix + "info" |
| 44 | +) |
| 45 | + |
| 46 | +// TODO: Refactor this logic, centralize it. See: https://github.com/open-telemetry/opentelemetry-operator/issues/2603 |
| 47 | +type components struct { |
| 48 | + receivers []string |
| 49 | + processors []string |
| 50 | + exporters []string |
| 51 | + extensions []string |
| 52 | + connectors []string |
| 53 | +} |
| 54 | + |
| 55 | +// Metrics hold all gauges for the different metrics related to the CRs |
| 56 | +// +kubebuilder:object:generate=false |
| 57 | +type Metrics struct { |
| 58 | + modeCounter metric.Int64UpDownCounter |
| 59 | + receiversCounter metric.Int64UpDownCounter |
| 60 | + exporterCounter metric.Int64UpDownCounter |
| 61 | + processorCounter metric.Int64UpDownCounter |
| 62 | + extensionsCounter metric.Int64UpDownCounter |
| 63 | + connectorsCounter metric.Int64UpDownCounter |
| 64 | +} |
| 65 | + |
| 66 | +// BootstrapMetrics configures the OpenTelemetry meter provider with the Prometheus exporter. |
| 67 | +func BootstrapMetrics() (metric.MeterProvider, error) { |
| 68 | + exporter, err := prometheus.New(prometheus.WithRegisterer(metrics.Registry)) |
| 69 | + if err != nil { |
| 70 | + return nil, err |
| 71 | + } |
| 72 | + return sdkmetric.NewMeterProvider(sdkmetric.WithReader(exporter)), err |
| 73 | +} |
| 74 | + |
| 75 | +func NewMetrics(prv metric.MeterProvider, ctx context.Context, cl client.Reader) (*Metrics, error) { |
| 76 | + meter := prv.Meter(meterName) |
| 77 | + modeCounter, err := meter.Int64UpDownCounter(modeMetricName) |
| 78 | + if err != nil { |
| 79 | + return nil, err |
| 80 | + } |
| 81 | + receiversCounter, err := meter.Int64UpDownCounter(receiversMetricName) |
| 82 | + if err != nil { |
| 83 | + return nil, err |
| 84 | + } |
| 85 | + |
| 86 | + exporterCounter, err := meter.Int64UpDownCounter(exportersMetricName) |
| 87 | + if err != nil { |
| 88 | + return nil, err |
| 89 | + } |
| 90 | + |
| 91 | + processorCounter, err := meter.Int64UpDownCounter(processorsMetricName) |
| 92 | + if err != nil { |
| 93 | + return nil, err |
| 94 | + } |
| 95 | + |
| 96 | + extensionsCounter, err := meter.Int64UpDownCounter(extensionsMetricName) |
| 97 | + if err != nil { |
| 98 | + return nil, err |
| 99 | + } |
| 100 | + |
| 101 | + connectorsCounter, err := meter.Int64UpDownCounter(connectorsMetricName) |
| 102 | + if err != nil { |
| 103 | + return nil, err |
| 104 | + } |
| 105 | + |
| 106 | + m := &Metrics{ |
| 107 | + modeCounter: modeCounter, |
| 108 | + receiversCounter: receiversCounter, |
| 109 | + exporterCounter: exporterCounter, |
| 110 | + processorCounter: processorCounter, |
| 111 | + extensionsCounter: extensionsCounter, |
| 112 | + connectorsCounter: connectorsCounter, |
| 113 | + } |
| 114 | + |
| 115 | + err = m.init(ctx, cl) |
| 116 | + if err != nil { |
| 117 | + return nil, err |
| 118 | + } |
| 119 | + return m, nil |
| 120 | +} |
| 121 | + |
| 122 | +// Init metrics from the first time the operator starts. |
| 123 | +func (m *Metrics) init(ctx context.Context, cl client.Reader) error { |
| 124 | + opts := []client.ListOption{ |
| 125 | + client.MatchingLabels(map[string]string{ |
| 126 | + "app.kubernetes.io/managed-by": "opentelemetry-operator", |
| 127 | + }), |
| 128 | + } |
| 129 | + list := &OpenTelemetryCollectorList{} |
| 130 | + if err := cl.List(ctx, list, opts...); err != nil { |
| 131 | + return fmt.Errorf("failed to list: %w", err) |
| 132 | + } |
| 133 | + |
| 134 | + for i := range list.Items { |
| 135 | + m.create(ctx, &list.Items[i]) |
| 136 | + } |
| 137 | + return nil |
| 138 | +} |
| 139 | + |
| 140 | +func (m *Metrics) create(ctx context.Context, collector *OpenTelemetryCollector) { |
| 141 | + m.updateComponentCounters(ctx, collector, true) |
| 142 | + m.updateGeneralCRMetricsComponents(ctx, collector, true) |
| 143 | +} |
| 144 | + |
| 145 | +func (m *Metrics) delete(ctx context.Context, collector *OpenTelemetryCollector) { |
| 146 | + m.updateComponentCounters(ctx, collector, false) |
| 147 | + m.updateGeneralCRMetricsComponents(ctx, collector, false) |
| 148 | +} |
| 149 | + |
| 150 | +func (m *Metrics) update(ctx context.Context, oldCollector *OpenTelemetryCollector, newCollector *OpenTelemetryCollector) { |
| 151 | + m.delete(ctx, oldCollector) |
| 152 | + m.create(ctx, newCollector) |
| 153 | +} |
| 154 | + |
| 155 | +func (m *Metrics) updateGeneralCRMetricsComponents(ctx context.Context, collector *OpenTelemetryCollector, up bool) { |
| 156 | + |
| 157 | + inc := 1 |
| 158 | + if !up { |
| 159 | + inc = -1 |
| 160 | + } |
| 161 | + m.modeCounter.Add(ctx, int64(inc), metric.WithAttributes( |
| 162 | + attribute.Key("collector_name").String(collector.Name), |
| 163 | + attribute.Key("namespace").String(collector.Namespace), |
| 164 | + attribute.Key("type").String(string(collector.Spec.Mode)), |
| 165 | + )) |
| 166 | +} |
| 167 | +func (m *Metrics) updateComponentCounters(ctx context.Context, collector *OpenTelemetryCollector, up bool) { |
| 168 | + components := getComponentsFromConfig(collector.Spec.Config) |
| 169 | + moveCounter(ctx, collector, components.receivers, m.receiversCounter, up) |
| 170 | + moveCounter(ctx, collector, components.exporters, m.exporterCounter, up) |
| 171 | + moveCounter(ctx, collector, components.processors, m.processorCounter, up) |
| 172 | + moveCounter(ctx, collector, components.extensions, m.extensionsCounter, up) |
| 173 | + moveCounter(ctx, collector, components.connectors, m.connectorsCounter, up) |
| 174 | + |
| 175 | +} |
| 176 | + |
| 177 | +func extractElements(elements map[string]interface{}) []string { |
| 178 | + // TODO: we should get rid of this method and centralize the parse logic |
| 179 | + // see https://github.com/open-telemetry/opentelemetry-operator/issues/2603 |
| 180 | + if elements == nil { |
| 181 | + return []string{} |
| 182 | + } |
| 183 | + |
| 184 | + itemsMap := map[string]struct{}{} |
| 185 | + var items []string |
| 186 | + for key := range elements { |
| 187 | + itemName := strings.SplitN(key, "/", 2)[0] |
| 188 | + itemsMap[itemName] = struct{}{} |
| 189 | + } |
| 190 | + for key := range itemsMap { |
| 191 | + items = append(items, key) |
| 192 | + } |
| 193 | + return items |
| 194 | +} |
| 195 | + |
| 196 | +func getComponentsFromConfig(yamlContent Config) *components { |
| 197 | + |
| 198 | + info := &components{ |
| 199 | + receivers: extractElements(yamlContent.Receivers.Object), |
| 200 | + exporters: extractElements(yamlContent.Exporters.Object), |
| 201 | + } |
| 202 | + |
| 203 | + if yamlContent.Processors != nil { |
| 204 | + info.processors = extractElements(yamlContent.Processors.Object) |
| 205 | + } |
| 206 | + |
| 207 | + if yamlContent.Extensions != nil { |
| 208 | + info.extensions = extractElements(yamlContent.Extensions.Object) |
| 209 | + } |
| 210 | + |
| 211 | + if yamlContent.Connectors != nil { |
| 212 | + info.connectors = extractElements(yamlContent.Connectors.Object) |
| 213 | + } |
| 214 | + |
| 215 | + return info |
| 216 | +} |
| 217 | + |
| 218 | +func moveCounter( |
| 219 | + ctx context.Context, collector *OpenTelemetryCollector, types []string, upDown metric.Int64UpDownCounter, up bool) { |
| 220 | + for _, exporter := range types { |
| 221 | + inc := 1 |
| 222 | + if !up { |
| 223 | + inc = -1 |
| 224 | + } |
| 225 | + upDown.Add(ctx, int64(inc), metric.WithAttributes( |
| 226 | + attribute.Key("collector_name").String(collector.Name), |
| 227 | + attribute.Key("namespace").String(collector.Namespace), |
| 228 | + attribute.Key("type").String(exporter), |
| 229 | + )) |
| 230 | + } |
| 231 | +} |
0 commit comments