khenaidoo | d948f77 | 2021-08-11 17:49:24 -0400 | [diff] [blame] | 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 | |
| 15 | package etcdserver |
| 16 | |
| 17 | import ( |
| 18 | goruntime "runtime" |
| 19 | "time" |
| 20 | |
| 21 | "github.com/coreos/etcd/pkg/runtime" |
| 22 | "github.com/coreos/etcd/version" |
| 23 | "github.com/prometheus/client_golang/prometheus" |
| 24 | ) |
| 25 | |
| 26 | var ( |
| 27 | hasLeader = prometheus.NewGauge(prometheus.GaugeOpts{ |
| 28 | Namespace: "etcd", |
| 29 | Subsystem: "server", |
| 30 | Name: "has_leader", |
| 31 | Help: "Whether or not a leader exists. 1 is existence, 0 is not.", |
| 32 | }) |
| 33 | isLeader = prometheus.NewGauge(prometheus.GaugeOpts{ |
| 34 | Namespace: "etcd", |
| 35 | Subsystem: "server", |
| 36 | Name: "is_leader", |
| 37 | Help: "Whether or not this member is a leader. 1 if is, 0 otherwise.", |
| 38 | }) |
| 39 | leaderChanges = prometheus.NewCounter(prometheus.CounterOpts{ |
| 40 | Namespace: "etcd", |
| 41 | Subsystem: "server", |
| 42 | Name: "leader_changes_seen_total", |
| 43 | Help: "The number of leader changes seen.", |
| 44 | }) |
| 45 | heartbeatSendFailures = prometheus.NewCounter(prometheus.CounterOpts{ |
| 46 | Namespace: "etcd", |
| 47 | Subsystem: "server", |
| 48 | Name: "heartbeat_send_failures_total", |
| 49 | Help: "The total number of leader heartbeat send failures (likely overloaded from slow disk).", |
| 50 | }) |
| 51 | slowApplies = prometheus.NewCounter(prometheus.CounterOpts{ |
| 52 | Namespace: "etcd", |
| 53 | Subsystem: "server", |
| 54 | Name: "slow_apply_total", |
| 55 | Help: "The total number of slow apply requests (likely overloaded from slow disk).", |
| 56 | }) |
| 57 | applySnapshotInProgress = prometheus.NewGauge(prometheus.GaugeOpts{ |
| 58 | Namespace: "etcd", |
| 59 | Subsystem: "server", |
| 60 | Name: "snapshot_apply_in_progress_total", |
| 61 | Help: "1 if the server is applying the incoming snapshot. 0 if none.", |
| 62 | }) |
| 63 | proposalsCommitted = prometheus.NewGauge(prometheus.GaugeOpts{ |
| 64 | Namespace: "etcd", |
| 65 | Subsystem: "server", |
| 66 | Name: "proposals_committed_total", |
| 67 | Help: "The total number of consensus proposals committed.", |
| 68 | }) |
| 69 | proposalsApplied = prometheus.NewGauge(prometheus.GaugeOpts{ |
| 70 | Namespace: "etcd", |
| 71 | Subsystem: "server", |
| 72 | Name: "proposals_applied_total", |
| 73 | Help: "The total number of consensus proposals applied.", |
| 74 | }) |
| 75 | proposalsPending = prometheus.NewGauge(prometheus.GaugeOpts{ |
| 76 | Namespace: "etcd", |
| 77 | Subsystem: "server", |
| 78 | Name: "proposals_pending", |
| 79 | Help: "The current number of pending proposals to commit.", |
| 80 | }) |
| 81 | proposalsFailed = prometheus.NewCounter(prometheus.CounterOpts{ |
| 82 | Namespace: "etcd", |
| 83 | Subsystem: "server", |
| 84 | Name: "proposals_failed_total", |
| 85 | Help: "The total number of failed proposals seen.", |
| 86 | }) |
| 87 | leaseExpired = prometheus.NewCounter(prometheus.CounterOpts{ |
| 88 | Namespace: "etcd_debugging", |
| 89 | Subsystem: "server", |
| 90 | Name: "lease_expired_total", |
| 91 | Help: "The total number of expired leases.", |
| 92 | }) |
| 93 | slowReadIndex = prometheus.NewCounter(prometheus.CounterOpts{ |
| 94 | Namespace: "etcd", |
| 95 | Subsystem: "server", |
| 96 | Name: "slow_read_indexes_total", |
| 97 | Help: "The total number of pending read indexes not in sync with leader's or timed out read index requests.", |
| 98 | }) |
| 99 | readIndexFailed = prometheus.NewCounter(prometheus.CounterOpts{ |
| 100 | Namespace: "etcd", |
| 101 | Subsystem: "server", |
| 102 | Name: "read_indexes_failed_total", |
| 103 | Help: "The total number of failed read indexes seen.", |
| 104 | }) |
| 105 | quotaBackendBytes = prometheus.NewGauge(prometheus.GaugeOpts{ |
| 106 | Namespace: "etcd", |
| 107 | Subsystem: "server", |
| 108 | Name: "quota_backend_bytes", |
| 109 | Help: "Current backend storage quota size in bytes.", |
| 110 | }) |
| 111 | currentVersion = prometheus.NewGaugeVec(prometheus.GaugeOpts{ |
| 112 | Namespace: "etcd", |
| 113 | Subsystem: "server", |
| 114 | Name: "version", |
| 115 | Help: "Which version is running. 1 for 'server_version' label with current version.", |
| 116 | }, |
| 117 | []string{"server_version"}) |
| 118 | currentGoVersion = prometheus.NewGaugeVec(prometheus.GaugeOpts{ |
| 119 | Namespace: "etcd", |
| 120 | Subsystem: "server", |
| 121 | Name: "go_version", |
| 122 | Help: "Which Go version server is running with. 1 for 'server_go_version' label with current version.", |
| 123 | }, |
| 124 | []string{"server_go_version"}) |
| 125 | serverID = prometheus.NewGaugeVec(prometheus.GaugeOpts{ |
| 126 | Namespace: "etcd", |
| 127 | Subsystem: "server", |
| 128 | Name: "id", |
| 129 | Help: "Server or member ID in hexadecimal format. 1 for 'server_id' label with current ID.", |
| 130 | }, |
| 131 | []string{"server_id"}) |
| 132 | |
| 133 | fdUsed = prometheus.NewGauge(prometheus.GaugeOpts{ |
| 134 | Namespace: "os", |
| 135 | Subsystem: "fd", |
| 136 | Name: "used", |
| 137 | Help: "The number of used file descriptors.", |
| 138 | }) |
| 139 | fdLimit = prometheus.NewGauge(prometheus.GaugeOpts{ |
| 140 | Namespace: "os", |
| 141 | Subsystem: "fd", |
| 142 | Name: "limit", |
| 143 | Help: "The file descriptor limit.", |
| 144 | }) |
| 145 | ) |
| 146 | |
| 147 | func init() { |
| 148 | prometheus.MustRegister(hasLeader) |
| 149 | prometheus.MustRegister(isLeader) |
| 150 | prometheus.MustRegister(leaderChanges) |
| 151 | prometheus.MustRegister(heartbeatSendFailures) |
| 152 | prometheus.MustRegister(slowApplies) |
| 153 | prometheus.MustRegister(applySnapshotInProgress) |
| 154 | prometheus.MustRegister(proposalsCommitted) |
| 155 | prometheus.MustRegister(proposalsApplied) |
| 156 | prometheus.MustRegister(proposalsPending) |
| 157 | prometheus.MustRegister(proposalsFailed) |
| 158 | prometheus.MustRegister(leaseExpired) |
| 159 | prometheus.MustRegister(slowReadIndex) |
| 160 | prometheus.MustRegister(readIndexFailed) |
| 161 | prometheus.MustRegister(quotaBackendBytes) |
| 162 | prometheus.MustRegister(currentVersion) |
| 163 | prometheus.MustRegister(currentGoVersion) |
| 164 | prometheus.MustRegister(serverID) |
| 165 | prometheus.MustRegister(fdUsed) |
| 166 | prometheus.MustRegister(fdLimit) |
| 167 | |
| 168 | currentVersion.With(prometheus.Labels{ |
| 169 | "server_version": version.Version, |
| 170 | }).Set(1) |
| 171 | currentGoVersion.With(prometheus.Labels{ |
| 172 | "server_go_version": goruntime.Version(), |
| 173 | }).Set(1) |
| 174 | } |
| 175 | |
| 176 | func monitorFileDescriptor(done <-chan struct{}) { |
| 177 | // This ticker will check File Descriptor Requirements ,and count all fds in used. |
| 178 | // And recorded some logs when in used >= limit/5*4. Just recorded message. |
| 179 | // If fds was more than 10K,It's low performance due to FDUsage() works. |
| 180 | // So need to increase it. |
| 181 | // See https://github.com/etcd-io/etcd/issues/11969 for more detail. |
| 182 | ticker := time.NewTicker(10 * time.Minute) |
| 183 | defer ticker.Stop() |
| 184 | for { |
| 185 | used, err := runtime.FDUsage() |
| 186 | if err != nil { |
| 187 | plog.Errorf("cannot monitor file descriptor usage (%v)", err) |
| 188 | return |
| 189 | } |
| 190 | fdUsed.Set(float64(used)) |
| 191 | limit, err := runtime.FDLimit() |
| 192 | if err != nil { |
| 193 | plog.Errorf("cannot monitor file descriptor usage (%v)", err) |
| 194 | return |
| 195 | } |
| 196 | fdLimit.Set(float64(limit)) |
| 197 | if used >= limit/5*4 { |
| 198 | plog.Warningf("80%% of the file descriptor limit is used [used = %d, limit = %d]", used, limit) |
| 199 | } |
| 200 | select { |
| 201 | case <-ticker.C: |
| 202 | case <-done: |
| 203 | return |
| 204 | } |
| 205 | } |
| 206 | } |