Girish Kumar | 46d7c3a | 2020-05-18 12:06:33 +0000 | [diff] [blame] | 1 | // Copyright (c) 2017 Uber Technologies, Inc. |
| 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 rpcmetrics |
| 16 | |
| 17 | import ( |
| 18 | "sync" |
| 19 | |
| 20 | "github.com/uber/jaeger-lib/metrics" |
| 21 | ) |
| 22 | |
| 23 | const ( |
| 24 | otherEndpointsPlaceholder = "other" |
| 25 | endpointNameMetricTag = "endpoint" |
| 26 | ) |
| 27 | |
| 28 | // Metrics is a collection of metrics for an endpoint describing |
| 29 | // throughput, success, errors, and performance. |
| 30 | type Metrics struct { |
| 31 | // RequestCountSuccess is a counter of the total number of successes. |
| 32 | RequestCountSuccess metrics.Counter `metric:"requests" tags:"error=false"` |
| 33 | |
| 34 | // RequestCountFailures is a counter of the number of times any failure has been observed. |
| 35 | RequestCountFailures metrics.Counter `metric:"requests" tags:"error=true"` |
| 36 | |
| 37 | // RequestLatencySuccess is a latency histogram of successful requests. |
| 38 | RequestLatencySuccess metrics.Timer `metric:"request_latency" tags:"error=false"` |
| 39 | |
| 40 | // RequestLatencyFailures is a latency histogram of failed requests. |
| 41 | RequestLatencyFailures metrics.Timer `metric:"request_latency" tags:"error=true"` |
| 42 | |
| 43 | // HTTPStatusCode2xx is a counter of the total number of requests with HTTP status code 200-299 |
| 44 | HTTPStatusCode2xx metrics.Counter `metric:"http_requests" tags:"status_code=2xx"` |
| 45 | |
| 46 | // HTTPStatusCode3xx is a counter of the total number of requests with HTTP status code 300-399 |
| 47 | HTTPStatusCode3xx metrics.Counter `metric:"http_requests" tags:"status_code=3xx"` |
| 48 | |
| 49 | // HTTPStatusCode4xx is a counter of the total number of requests with HTTP status code 400-499 |
| 50 | HTTPStatusCode4xx metrics.Counter `metric:"http_requests" tags:"status_code=4xx"` |
| 51 | |
| 52 | // HTTPStatusCode5xx is a counter of the total number of requests with HTTP status code 500-599 |
| 53 | HTTPStatusCode5xx metrics.Counter `metric:"http_requests" tags:"status_code=5xx"` |
| 54 | } |
| 55 | |
| 56 | func (m *Metrics) recordHTTPStatusCode(statusCode uint16) { |
| 57 | if statusCode >= 200 && statusCode < 300 { |
| 58 | m.HTTPStatusCode2xx.Inc(1) |
| 59 | } else if statusCode >= 300 && statusCode < 400 { |
| 60 | m.HTTPStatusCode3xx.Inc(1) |
| 61 | } else if statusCode >= 400 && statusCode < 500 { |
| 62 | m.HTTPStatusCode4xx.Inc(1) |
| 63 | } else if statusCode >= 500 && statusCode < 600 { |
| 64 | m.HTTPStatusCode5xx.Inc(1) |
| 65 | } |
| 66 | } |
| 67 | |
| 68 | // MetricsByEndpoint is a registry/cache of metrics for each unique endpoint name. |
| 69 | // Only maxNumberOfEndpoints Metrics are stored, all other endpoint names are mapped |
| 70 | // to a generic endpoint name "other". |
| 71 | type MetricsByEndpoint struct { |
| 72 | metricsFactory metrics.Factory |
| 73 | endpoints *normalizedEndpoints |
| 74 | metricsByEndpoint map[string]*Metrics |
| 75 | mux sync.RWMutex |
| 76 | } |
| 77 | |
| 78 | func newMetricsByEndpoint( |
| 79 | metricsFactory metrics.Factory, |
| 80 | normalizer NameNormalizer, |
| 81 | maxNumberOfEndpoints int, |
| 82 | ) *MetricsByEndpoint { |
| 83 | return &MetricsByEndpoint{ |
| 84 | metricsFactory: metricsFactory, |
| 85 | endpoints: newNormalizedEndpoints(maxNumberOfEndpoints, normalizer), |
| 86 | metricsByEndpoint: make(map[string]*Metrics, maxNumberOfEndpoints+1), // +1 for "other" |
| 87 | } |
| 88 | } |
| 89 | |
| 90 | func (m *MetricsByEndpoint) get(endpoint string) *Metrics { |
| 91 | safeName := m.endpoints.normalize(endpoint) |
| 92 | if safeName == "" { |
| 93 | safeName = otherEndpointsPlaceholder |
| 94 | } |
| 95 | m.mux.RLock() |
| 96 | met := m.metricsByEndpoint[safeName] |
| 97 | m.mux.RUnlock() |
| 98 | if met != nil { |
| 99 | return met |
| 100 | } |
| 101 | |
| 102 | return m.getWithWriteLock(safeName) |
| 103 | } |
| 104 | |
| 105 | // split to make easier to test |
| 106 | func (m *MetricsByEndpoint) getWithWriteLock(safeName string) *Metrics { |
| 107 | m.mux.Lock() |
| 108 | defer m.mux.Unlock() |
| 109 | |
| 110 | // it is possible that the name has been already registered after we released |
| 111 | // the read lock and before we grabbed the write lock, so check for that. |
| 112 | if met, ok := m.metricsByEndpoint[safeName]; ok { |
| 113 | return met |
| 114 | } |
| 115 | |
| 116 | // it would be nice to create the struct before locking, since Init() is somewhat |
| 117 | // expensive, however some metrics backends (e.g. expvar) may not like duplicate metrics. |
| 118 | met := &Metrics{} |
| 119 | tags := map[string]string{endpointNameMetricTag: safeName} |
| 120 | metrics.Init(met, m.metricsFactory, tags) |
| 121 | |
| 122 | m.metricsByEndpoint[safeName] = met |
| 123 | return met |
| 124 | } |