k8s.io/apiserver@v0.31.1/pkg/server/egressselector/metrics/metrics.go (about)

     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 metrics
    18  
    19  import (
    20  	"time"
    21  
    22  	"k8s.io/component-base/metrics"
    23  	"k8s.io/component-base/metrics/legacyregistry"
    24  	"k8s.io/utils/clock"
    25  )
    26  
    27  const (
    28  	namespace = "apiserver"
    29  	subsystem = "egress_dialer"
    30  
    31  	// ProtocolHTTPConnect means that the proxy protocol is http-connect.
    32  	ProtocolHTTPConnect = "http_connect"
    33  	// ProtocolGRPC means that the proxy protocol is the GRPC protocol.
    34  	ProtocolGRPC = "grpc"
    35  	// TransportTCP means that the transport is TCP.
    36  	TransportTCP = "tcp"
    37  	// TransportUDS means that the transport is UDS.
    38  	TransportUDS = "uds"
    39  	// StageConnect indicates that the dial failed at establishing connection to the proxy server.
    40  	StageConnect = "connect"
    41  	// StageProxy indicates that the dial failed at requesting the proxy server to proxy.
    42  	StageProxy = "proxy"
    43  )
    44  
    45  var (
    46  	// Use buckets ranging from 5 ms to 12.5 seconds.
    47  	latencyBuckets = []float64{0.005, 0.025, 0.1, 0.5, 2.5, 12.5}
    48  
    49  	// Metrics provides access to all dial metrics.
    50  	Metrics = newDialMetrics()
    51  )
    52  
    53  // DialMetrics instruments dials to proxy server with prometheus metrics.
    54  type DialMetrics struct {
    55  	clock     clock.Clock
    56  	starts    *metrics.CounterVec
    57  	latencies *metrics.HistogramVec
    58  	failures  *metrics.CounterVec
    59  }
    60  
    61  // newDialMetrics create a new DialMetrics, configured with default metric names.
    62  func newDialMetrics() *DialMetrics {
    63  	starts := metrics.NewCounterVec(
    64  		&metrics.CounterOpts{
    65  			Namespace:      namespace,
    66  			Subsystem:      subsystem,
    67  			Name:           "dial_start_total",
    68  			Help:           "Dial starts, labeled by the protocol (http-connect or grpc) and transport (tcp or uds).",
    69  			StabilityLevel: metrics.ALPHA,
    70  		},
    71  		[]string{"protocol", "transport"},
    72  	)
    73  
    74  	latencies := metrics.NewHistogramVec(
    75  		&metrics.HistogramOpts{
    76  			Namespace:      namespace,
    77  			Subsystem:      subsystem,
    78  			Name:           "dial_duration_seconds",
    79  			Help:           "Dial latency histogram in seconds, labeled by the protocol (http-connect or grpc), transport (tcp or uds)",
    80  			Buckets:        latencyBuckets,
    81  			StabilityLevel: metrics.ALPHA,
    82  		},
    83  		[]string{"protocol", "transport"},
    84  	)
    85  
    86  	failures := metrics.NewCounterVec(
    87  		&metrics.CounterOpts{
    88  			Namespace:      namespace,
    89  			Subsystem:      subsystem,
    90  			Name:           "dial_failure_count",
    91  			Help:           "Dial failure count, labeled by the protocol (http-connect or grpc), transport (tcp or uds), and stage (connect or proxy). The stage indicates at which stage the dial failed",
    92  			StabilityLevel: metrics.ALPHA,
    93  		},
    94  		[]string{"protocol", "transport", "stage"},
    95  	)
    96  
    97  	legacyregistry.MustRegister(starts)
    98  	legacyregistry.MustRegister(latencies)
    99  	legacyregistry.MustRegister(failures)
   100  	return &DialMetrics{starts: starts, latencies: latencies, failures: failures, clock: clock.RealClock{}}
   101  }
   102  
   103  // Clock returns the clock.
   104  func (m *DialMetrics) Clock() clock.Clock {
   105  	return m.clock
   106  }
   107  
   108  // SetClock sets the clock.
   109  func (m *DialMetrics) SetClock(c clock.Clock) {
   110  	m.clock = c
   111  }
   112  
   113  // Reset resets the metrics.
   114  func (m *DialMetrics) Reset() {
   115  	m.starts.Reset()
   116  	m.latencies.Reset()
   117  	m.failures.Reset()
   118  }
   119  
   120  // ObserveDialStart records the start of a dial attempt, labeled by protocol, transport.
   121  func (m *DialMetrics) ObserveDialStart(protocol, transport string) {
   122  	m.starts.WithLabelValues(protocol, transport).Inc()
   123  }
   124  
   125  // ObserveDialLatency records the latency of a dial, labeled by protocol, transport.
   126  func (m *DialMetrics) ObserveDialLatency(elapsed time.Duration, protocol, transport string) {
   127  	m.latencies.WithLabelValues(protocol, transport).Observe(elapsed.Seconds())
   128  }
   129  
   130  // ObserveDialFailure records a failed dial, labeled by protocol, transport, and the stage the dial failed at.
   131  func (m *DialMetrics) ObserveDialFailure(protocol, transport, stage string) {
   132  	m.failures.WithLabelValues(protocol, transport, stage).Inc()
   133  }