sigs.k8s.io/cluster-api@v1.7.1/internal/runtime/metrics/metrics.go (about)

     1  /*
     2  Copyright 2022 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 metrics provides functions for creating Runtime SDK related metrics.
    18  package metrics
    19  
    20  import (
    21  	"net/http"
    22  	"net/url"
    23  	"strconv"
    24  	"time"
    25  
    26  	"github.com/prometheus/client_golang/prometheus"
    27  	"k8s.io/apimachinery/pkg/runtime"
    28  	ctrlmetrics "sigs.k8s.io/controller-runtime/pkg/metrics"
    29  
    30  	runtimecatalog "sigs.k8s.io/cluster-api/exp/runtime/catalog"
    31  	runtimehooksv1 "sigs.k8s.io/cluster-api/exp/runtime/hooks/api/v1alpha1"
    32  )
    33  
    34  func init() {
    35  	// Register the metrics at the controller-runtime metrics registry.
    36  	ctrlmetrics.Registry.MustRegister(RequestsTotal.metric)
    37  	ctrlmetrics.Registry.MustRegister(RequestDuration.metric)
    38  }
    39  
    40  // Metrics subsystem and all of the keys used by the Runtime SDK.
    41  const (
    42  	runtimeSDKSubsystem   = "capi_runtime_sdk"
    43  	unknownResponseStatus = "Unknown"
    44  )
    45  
    46  var (
    47  	// RequestsTotal reports request results.
    48  	RequestsTotal = requestsTotalObserver{
    49  		prometheus.NewCounterVec(prometheus.CounterOpts{
    50  			Subsystem: runtimeSDKSubsystem,
    51  			Name:      "requests_total",
    52  			Help:      "Number of HTTP requests, partitioned by status code, host, hook and response status.",
    53  		}, []string{"code", "host", "group", "version", "hook", "status"}),
    54  	}
    55  	// RequestDuration reports the request latency in seconds.
    56  	RequestDuration = requestDurationObserver{
    57  		prometheus.NewHistogramVec(prometheus.HistogramOpts{
    58  			Subsystem: runtimeSDKSubsystem,
    59  			Name:      "request_duration_seconds",
    60  			Help:      "Request duration in seconds, broken down by hook and host.",
    61  			Buckets: []float64{0.005, 0.025, 0.05, 0.1, 0.2, 0.4, 0.6, 0.8, 1.0, 1.25, 1.5, 2, 3,
    62  				4, 5, 6, 8, 10, 15, 20, 30, 45, 60},
    63  		}, []string{"host", "group", "version", "hook"}),
    64  	}
    65  )
    66  
    67  type requestsTotalObserver struct {
    68  	metric *prometheus.CounterVec
    69  }
    70  
    71  // Observe observes a http request result and increments the metric for the given
    72  // http status code, host, gvh and response.
    73  func (m *requestsTotalObserver) Observe(req *http.Request, resp *http.Response, gvh runtimecatalog.GroupVersionHook, err error, response runtime.Object) {
    74  	host := req.URL.Host
    75  
    76  	// Errors can be arbitrary strings. Unbound label cardinality is not suitable for a metric
    77  	// system so they are reported as `<error>`.
    78  	code := "<error>"
    79  	if err == nil {
    80  		code = strconv.Itoa(resp.StatusCode)
    81  	}
    82  
    83  	status := unknownResponseStatus
    84  	if responseObject, ok := response.(runtimehooksv1.ResponseObject); ok && responseObject.GetStatus() != "" {
    85  		status = string(responseObject.GetStatus())
    86  	}
    87  
    88  	m.metric.WithLabelValues(code, host, gvh.Group, gvh.Version, gvh.Hook, status).Inc()
    89  }
    90  
    91  type requestDurationObserver struct {
    92  	metric *prometheus.HistogramVec
    93  }
    94  
    95  // Observe increments the request latency metric for the given host and gvh.
    96  func (m *requestDurationObserver) Observe(gvh runtimecatalog.GroupVersionHook, u url.URL, latency time.Duration) {
    97  	m.metric.WithLabelValues(u.Host, gvh.Group, gvh.Version, gvh.Hook).Observe(latency.Seconds())
    98  }