metrics.go 6.1 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191
  1. // Copyright 2015 The etcd 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. package etcdserver
  15. import (
  16. goruntime "runtime"
  17. "time"
  18. "go.etcd.io/etcd/pkg/runtime"
  19. "go.etcd.io/etcd/version"
  20. "github.com/prometheus/client_golang/prometheus"
  21. "go.uber.org/zap"
  22. )
  23. var (
  24. hasLeader = prometheus.NewGauge(prometheus.GaugeOpts{
  25. Namespace: "etcd",
  26. Subsystem: "server",
  27. Name: "has_leader",
  28. Help: "Whether or not a leader exists. 1 is existence, 0 is not.",
  29. })
  30. isLeader = prometheus.NewGauge(prometheus.GaugeOpts{
  31. Namespace: "etcd",
  32. Subsystem: "server",
  33. Name: "is_leader",
  34. Help: "Whether or not this member is a leader. 1 if is, 0 otherwise.",
  35. })
  36. leaderChanges = prometheus.NewCounter(prometheus.CounterOpts{
  37. Namespace: "etcd",
  38. Subsystem: "server",
  39. Name: "leader_changes_seen_total",
  40. Help: "The number of leader changes seen.",
  41. })
  42. heartbeatSendFailures = prometheus.NewCounter(prometheus.CounterOpts{
  43. Namespace: "etcd",
  44. Subsystem: "server",
  45. Name: "heartbeat_send_failures_total",
  46. Help: "The total number of leader heartbeat send failures (likely overloaded from slow disk).",
  47. })
  48. slowApplies = prometheus.NewCounter(prometheus.CounterOpts{
  49. Namespace: "etcd",
  50. Subsystem: "server",
  51. Name: "slow_apply_total",
  52. Help: "The total number of slow apply requests (likely overloaded from slow disk).",
  53. })
  54. proposalsCommitted = prometheus.NewGauge(prometheus.GaugeOpts{
  55. Namespace: "etcd",
  56. Subsystem: "server",
  57. Name: "proposals_committed_total",
  58. Help: "The total number of consensus proposals committed.",
  59. })
  60. proposalsApplied = prometheus.NewGauge(prometheus.GaugeOpts{
  61. Namespace: "etcd",
  62. Subsystem: "server",
  63. Name: "proposals_applied_total",
  64. Help: "The total number of consensus proposals applied.",
  65. })
  66. proposalsPending = prometheus.NewGauge(prometheus.GaugeOpts{
  67. Namespace: "etcd",
  68. Subsystem: "server",
  69. Name: "proposals_pending",
  70. Help: "The current number of pending proposals to commit.",
  71. })
  72. proposalsFailed = prometheus.NewCounter(prometheus.CounterOpts{
  73. Namespace: "etcd",
  74. Subsystem: "server",
  75. Name: "proposals_failed_total",
  76. Help: "The total number of failed proposals seen.",
  77. })
  78. slowReadIndex = prometheus.NewCounter(prometheus.CounterOpts{
  79. Namespace: "etcd",
  80. Subsystem: "server",
  81. Name: "slow_read_indexes_total",
  82. Help: "The total number of pending read indexes not in sync with leader's or timed out read index requests.",
  83. })
  84. readIndexFailed = prometheus.NewCounter(prometheus.CounterOpts{
  85. Namespace: "etcd",
  86. Subsystem: "server",
  87. Name: "read_indexes_failed_total",
  88. Help: "The total number of failed read indexes seen.",
  89. })
  90. leaseExpired = prometheus.NewCounter(prometheus.CounterOpts{
  91. Namespace: "etcd_debugging",
  92. Subsystem: "server",
  93. Name: "lease_expired_total",
  94. Help: "The total number of expired leases.",
  95. })
  96. quotaBackendBytes = prometheus.NewGauge(prometheus.GaugeOpts{
  97. Namespace: "etcd",
  98. Subsystem: "server",
  99. Name: "quota_backend_bytes",
  100. Help: "Current backend storage quota size in bytes.",
  101. })
  102. currentVersion = prometheus.NewGaugeVec(prometheus.GaugeOpts{
  103. Namespace: "etcd",
  104. Subsystem: "server",
  105. Name: "version",
  106. Help: "Which version is running. 1 for 'server_version' label with current version.",
  107. },
  108. []string{"server_version"})
  109. currentGoVersion = prometheus.NewGaugeVec(prometheus.GaugeOpts{
  110. Namespace: "etcd",
  111. Subsystem: "server",
  112. Name: "go_version",
  113. Help: "Which Go version server is running with. 1 for 'server_go_version' label with current version.",
  114. },
  115. []string{"server_go_version"})
  116. serverID = prometheus.NewGaugeVec(prometheus.GaugeOpts{
  117. Namespace: "etcd",
  118. Subsystem: "server",
  119. Name: "id",
  120. Help: "Server or member ID in hexadecimal format. 1 for 'server_id' label with current ID.",
  121. },
  122. []string{"server_id"})
  123. )
  124. func init() {
  125. prometheus.MustRegister(hasLeader)
  126. prometheus.MustRegister(isLeader)
  127. prometheus.MustRegister(leaderChanges)
  128. prometheus.MustRegister(heartbeatSendFailures)
  129. prometheus.MustRegister(slowApplies)
  130. prometheus.MustRegister(proposalsCommitted)
  131. prometheus.MustRegister(proposalsApplied)
  132. prometheus.MustRegister(proposalsPending)
  133. prometheus.MustRegister(proposalsFailed)
  134. prometheus.MustRegister(slowReadIndex)
  135. prometheus.MustRegister(readIndexFailed)
  136. prometheus.MustRegister(leaseExpired)
  137. prometheus.MustRegister(quotaBackendBytes)
  138. prometheus.MustRegister(currentVersion)
  139. prometheus.MustRegister(currentGoVersion)
  140. prometheus.MustRegister(serverID)
  141. currentVersion.With(prometheus.Labels{
  142. "server_version": version.Version,
  143. }).Set(1)
  144. currentGoVersion.With(prometheus.Labels{
  145. "server_go_version": goruntime.Version(),
  146. }).Set(1)
  147. }
  148. func monitorFileDescriptor(lg *zap.Logger, done <-chan struct{}) {
  149. ticker := time.NewTicker(5 * time.Second)
  150. defer ticker.Stop()
  151. for {
  152. used, err := runtime.FDUsage()
  153. if err != nil {
  154. if lg != nil {
  155. lg.Warn("failed to get file descriptor usage", zap.Error(err))
  156. } else {
  157. plog.Errorf("cannot monitor file descriptor usage (%v)", err)
  158. }
  159. return
  160. }
  161. limit, err := runtime.FDLimit()
  162. if err != nil {
  163. if lg != nil {
  164. lg.Warn("failed to get file descriptor limit", zap.Error(err))
  165. } else {
  166. plog.Errorf("cannot monitor file descriptor usage (%v)", err)
  167. }
  168. return
  169. }
  170. if used >= limit/5*4 {
  171. if lg != nil {
  172. lg.Warn("80%% of file descriptors are used", zap.Uint64("used", used), zap.Uint64("limit", limit))
  173. } else {
  174. plog.Warningf("80%% of the file descriptor limit is used [used = %d, limit = %d]", used, limit)
  175. }
  176. }
  177. select {
  178. case <-ticker.C:
  179. case <-done:
  180. return
  181. }
  182. }
  183. }