| 1 |
/* |
| 2 |
Copyright 2019 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 |
"context" |
| 21 |
"strconv" |
| 22 |
"strings" |
| 23 |
"sync" |
| 24 |
"time" |
| 25 |
|
| 26 |
epmetrics "k8s.io/apiserver/pkg/endpoints/metrics" |
| 27 |
apirequest "k8s.io/apiserver/pkg/endpoints/request" |
| 28 |
compbasemetrics "k8s.io/component-base/metrics" |
| 29 |
"k8s.io/component-base/metrics/legacyregistry" |
| 30 |
basemetricstestutil "k8s.io/component-base/metrics/testutil" |
| 31 |
) |
| 32 |
|
| 33 |
const ( |
| 34 |
namespace = "apiserver" |
| 35 |
subsystem = "flowcontrol" |
| 36 |
) |
| 37 |
|
| 38 |
const ( |
| 39 |
requestKind = "request_kind" |
| 40 |
priorityLevel = "priority_level" |
| 41 |
flowSchema = "flow_schema" |
| 42 |
phase = "phase" |
| 43 |
LabelNamePhase = "phase" |
| 44 |
LabelValueWaiting = "waiting" |
| 45 |
LabelValueExecuting = "executing" |
| 46 |
) |
| 47 |
|
| 48 |
var ( |
| 49 |
queueLengthBuckets = []float64{0, 10, 25, 50, 100, 250, 500, 1000} |
| 50 |
requestDurationSecondsBuckets = []float64{0, 0.005, 0.02, 0.05, 0.1, 0.2, 0.5, 1, 2, 5, 10, 15, 30} |
| 51 |
) |
| 52 |
|
| 53 |
var registerMetrics sync.Once |
| 54 |
|
| 55 |
// Register all metrics. |
| 56 |
func Register() { |
| 57 |
registerMetrics.Do(func() { |
| 58 |
for _, metric := range metrics { |
| 59 |
legacyregistry.MustRegister(metric) |
| 60 |
} |
| 61 |
}) |
| 62 |
} |
| 63 |
|
| 64 |
type resettable interface { |
| 65 |
Reset() |
| 66 |
} |
| 67 |
|
| 68 |
// Reset all resettable metrics to zero |
| 69 |
func Reset() { |
| 70 |
for _, metric := range metrics { |
| 71 |
if rm, ok := metric.(resettable); ok { |
| 72 |
rm.Reset() |
| 73 |
} |
| 74 |
} |
| 75 |
} |
| 76 |
|
| 77 |
// GatherAndCompare the given metrics with the given Prometheus syntax expected value |
| 78 |
func GatherAndCompare(expected string, metricNames ...string) error { |
| 79 |
return basemetricstestutil.GatherAndCompare(legacyregistry.DefaultGatherer, strings.NewReader(expected), metricNames...) |
| 80 |
} |
| 81 |
|
| 82 |
// Registerables is a slice of Registerable |
| 83 |
type Registerables []compbasemetrics.Registerable |
| 84 |
|
| 85 |
// Append adds more |
| 86 |
func (rs Registerables) Append(more ...compbasemetrics.Registerable) Registerables { |
| 87 |
return append(rs, more...) |
| 88 |
} |
| 89 |
|
| 90 |
var ( |
| 91 |
apiserverRejectedRequestsTotal = compbasemetrics.NewCounterVec( |
| 92 |
&compbasemetrics.CounterOpts{ |
| 93 |
Namespace: namespace, |
| 94 |
Subsystem: subsystem, |
| 95 |
Name: "rejected_requests_total", |
| 96 |
Help: "Number of requests rejected by API Priority and Fairness subsystem", |
| 97 |
StabilityLevel: compbasemetrics.BETA, |
| 98 |
}, |
| 99 |
[]string{priorityLevel, flowSchema, "reason"}, |
| 100 |
) |
| 101 |
apiserverDispatchedRequestsTotal = compbasemetrics.NewCounterVec( |
| 102 |
&compbasemetrics.CounterOpts{ |
| 103 |
Namespace: namespace, |
| 104 |
Subsystem: subsystem, |
| 105 |
Name: "dispatched_requests_total", |
| 106 |
Help: "Number of requests executed by API Priority and Fairness subsystem", |
| 107 |
StabilityLevel: compbasemetrics.BETA, |
| 108 |
}, |
| 109 |
[]string{priorityLevel, flowSchema}, |
| 110 |
) |
| 111 |
// PriorityLevelExecutionSeatsGaugeVec creates observers of seats occupied throughout execution for priority levels |
| 112 |
PriorityLevelExecutionSeatsGaugeVec = NewTimingRatioHistogramVec( |
| 113 |
&compbasemetrics.TimingHistogramOpts{ |
| 114 |
Namespace: namespace, |
| 115 |
Subsystem: subsystem, |
| 116 |
Name: "priority_level_seat_utilization", |
| 117 |
Help: "Observations, at the end of every nanosecond, of utilization of seats for any stage of execution (but only initial stage for WATCHes)", |
| 118 |
// Buckets for both 0.99 and 1.0 mean PromQL's histogram_quantile will reveal saturation |
| 119 |
Buckets: []float64{0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.95, 0.99, 1}, |
| 120 |
ConstLabels: map[string]string{phase: "executing"}, |
| 121 |
StabilityLevel: compbasemetrics.ALPHA, |
| 122 |
}, |
| 123 |
priorityLevel, |
| 124 |
) |
| 125 |
// PriorityLevelConcurrencyGaugeVec creates gauges of concurrency broken down by phase, priority level |
| 126 |
PriorityLevelConcurrencyGaugeVec = NewTimingRatioHistogramVec( |
| 127 |
&compbasemetrics.TimingHistogramOpts{ |
| 128 |
Namespace: namespace, |
| 129 |
Subsystem: subsystem, |
| 130 |
Name: "priority_level_request_utilization", |
| 131 |
Help: "Observations, at the end of every nanosecond, of number of requests (as a fraction of the relevant limit) waiting or in any stage of execution (but only initial stage for WATCHes)", |
| 132 |
// For executing: the denominator will be seats, so this metric will skew low. |
| 133 |
// For waiting: total queue capacity is generally quite generous, so this metric will skew low. |
| 134 |
Buckets: []float64{0, 0.001, 0.003, 0.01, 0.03, 0.1, 0.25, 0.5, 0.75, 1}, |
| 135 |
StabilityLevel: compbasemetrics.ALPHA, |
| 136 |
}, |
| 137 |
LabelNamePhase, priorityLevel, |
| 138 |
) |
| 139 |
// readWriteConcurrencyGaugeVec creates ratioed gauges of requests/limit broken down by phase and mutating vs readonly |
| 140 |
readWriteConcurrencyGaugeVec = NewTimingRatioHistogramVec( |
| 141 |
&compbasemetrics.TimingHistogramOpts{ |
| 142 |
Namespace: namespace, |
| 143 |
Subsystem: subsystem, |
| 144 |
Name: "read_vs_write_current_requests", |
| 145 |
Help: "Observations, at the end of every nanosecond, of the number of requests (as a fraction of the relevant limit) waiting or in regular stage of execution", |
| 146 |
// This metric will skew low for the same reason as the priority level metrics |
| 147 |
// and also because APF has a combined limit for mutating and readonly. |
| 148 |
Buckets: []float64{0, 0.001, 0.01, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.95, 0.99, 1}, |
| 149 |
StabilityLevel: compbasemetrics.ALPHA, |
| 150 |
}, |
| 151 |
LabelNamePhase, requestKind, |
| 152 |
) |
| 153 |
apiserverCurrentR = compbasemetrics.NewGaugeVec( |
| 154 |
&compbasemetrics.GaugeOpts{ |
| 155 |
Namespace: namespace, |
| 156 |
Subsystem: subsystem, |
| 157 |
Name: "current_r", |
| 158 |
Help: "R(time of last change)", |
| 159 |
StabilityLevel: compbasemetrics.ALPHA, |
| 160 |
}, |
| 161 |
[]string{priorityLevel}, |
| 162 |
) |
| 163 |
apiserverDispatchR = compbasemetrics.NewGaugeVec( |
| 164 |
&compbasemetrics.GaugeOpts{ |
| 165 |
Namespace: namespace, |
| 166 |
Subsystem: subsystem, |
| 167 |
Name: "dispatch_r", |
| 168 |
Help: "R(time of last dispatch)", |
| 169 |
StabilityLevel: compbasemetrics.ALPHA, |
| 170 |
}, |
| 171 |
[]string{priorityLevel}, |
| 172 |
) |
| 173 |
apiserverLatestS = compbasemetrics.NewGaugeVec( |
| 174 |
&compbasemetrics.GaugeOpts{ |
| 175 |
Namespace: namespace, |
| 176 |
Subsystem: subsystem, |
| 177 |
Name: "latest_s", |
| 178 |
Help: "S(most recently dispatched request)", |
| 179 |
StabilityLevel: compbasemetrics.ALPHA, |
| 180 |
}, |
| 181 |
[]string{priorityLevel}, |
| 182 |
) |
| 183 |
apiserverNextSBounds = compbasemetrics.NewGaugeVec( |
| 184 |
&compbasemetrics.GaugeOpts{ |
| 185 |
Namespace: namespace, |
| 186 |
Subsystem: subsystem, |
| 187 |
Name: "next_s_bounds", |
| 188 |
Help: "min and max, over queues, of S(oldest waiting request in queue)", |
| 189 |
StabilityLevel: compbasemetrics.ALPHA, |
| 190 |
}, |
| 191 |
[]string{priorityLevel, "bound"}, |
| 192 |
) |
| 193 |
apiserverNextDiscountedSBounds = compbasemetrics.NewGaugeVec( |
| 194 |
&compbasemetrics.GaugeOpts{ |
| 195 |
Namespace: namespace, |
| 196 |
Subsystem: subsystem, |
| 197 |
Name: "next_discounted_s_bounds", |
| 198 |
Help: "min and max, over queues, of S(oldest waiting request in queue) - estimated work in progress", |
| 199 |
StabilityLevel: compbasemetrics.ALPHA, |
| 200 |
}, |
| 201 |
[]string{priorityLevel, "bound"}, |
| 202 |
) |
| 203 |
apiserverCurrentInqueueRequests = compbasemetrics.NewGaugeVec( |
| 204 |
&compbasemetrics.GaugeOpts{ |
| 205 |
Namespace: namespace, |
| 206 |
Subsystem: subsystem, |
| 207 |
Name: "current_inqueue_requests", |
| 208 |
Help: "Number of requests currently pending in queues of the API Priority and Fairness subsystem", |
| 209 |
StabilityLevel: compbasemetrics.BETA, |
| 210 |
}, |
| 211 |
[]string{priorityLevel, flowSchema}, |
| 212 |
) |
| 213 |
apiserverCurrentInqueueSeats = compbasemetrics.NewGaugeVec( |
| 214 |
&compbasemetrics.GaugeOpts{ |
| 215 |
Namespace: namespace, |
| 216 |
Subsystem: subsystem, |
| 217 |
Name: "current_inqueue_seats", |
| 218 |
Help: "Number of seats currently pending in queues of the API Priority and Fairness subsystem", |
| 219 |
StabilityLevel: compbasemetrics.ALPHA, |
| 220 |
}, |
| 221 |
[]string{priorityLevel, flowSchema}, |
| 222 |
) |
| 223 |
apiserverRequestQueueLength = compbasemetrics.NewHistogramVec( |
| 224 |
&compbasemetrics.HistogramOpts{ |
| 225 |
Namespace: namespace, |
| 226 |
Subsystem: subsystem, |
| 227 |
Name: "request_queue_length_after_enqueue", |
| 228 |
Help: "Length of queue in the API Priority and Fairness subsystem, as seen by each request after it is enqueued", |
| 229 |
Buckets: queueLengthBuckets, |
| 230 |
StabilityLevel: compbasemetrics.ALPHA, |
| 231 |
}, |
| 232 |
[]string{priorityLevel, flowSchema}, |
| 233 |
) |
| 234 |
apiserverRequestConcurrencyLimit = compbasemetrics.NewGaugeVec( |
| 235 |
&compbasemetrics.GaugeOpts{ |
| 236 |
Namespace: namespace, |
| 237 |
Subsystem: subsystem, |
| 238 |
Name: "request_concurrency_limit", |
| 239 |
Help: "Nominal number of execution seats configured for each priority level", |
| 240 |
// Remove this metric once all suppported releases have the equal nominal_limit_seats metric |
| 241 |
DeprecatedVersion: "1.30.0", |
| 242 |
StabilityLevel: compbasemetrics.ALPHA, |
| 243 |
}, |
| 244 |
[]string{priorityLevel}, |
| 245 |
) |
| 246 |
apiserverCurrentExecutingRequests = compbasemetrics.NewGaugeVec( |
| 247 |
&compbasemetrics.GaugeOpts{ |
| 248 |
Namespace: namespace, |
| 249 |
Subsystem: subsystem, |
| 250 |
Name: "current_executing_requests", |
| 251 |
Help: "Number of requests in initial (for a WATCH) or any (for a non-WATCH) execution stage in the API Priority and Fairness subsystem", |
| 252 |
StabilityLevel: compbasemetrics.BETA, |
| 253 |
}, |
| 254 |
[]string{priorityLevel, flowSchema}, |
| 255 |
) |
| 256 |
apiserverCurrentExecutingSeats = compbasemetrics.NewGaugeVec( |
| 257 |
&compbasemetrics.GaugeOpts{ |
| 258 |
Namespace: namespace, |
| 259 |
Subsystem: subsystem, |
| 260 |
Name: "current_executing_seats", |
| 261 |
Help: "Concurrency (number of seats) occupied by the currently executing (initial stage for a WATCH, any stage otherwise) requests in the API Priority and Fairness subsystem", |
| 262 |
StabilityLevel: compbasemetrics.BETA, |
| 263 |
}, |
| 264 |
[]string{priorityLevel, flowSchema}, |
| 265 |
) |
| 266 |
apiserverRequestConcurrencyInUse = compbasemetrics.NewGaugeVec( |
| 267 |
&compbasemetrics.GaugeOpts{ |
| 268 |
Namespace: namespace, |
| 269 |
Subsystem: subsystem, |
| 270 |
Name: "request_concurrency_in_use", |
| 271 |
Help: "Concurrency (number of seats) occupied by the currently executing (initial stage for a WATCH, any stage otherwise) requests in the API Priority and Fairness subsystem", |
| 272 |
// Remove this metric once all suppported releases have the equal current_executing_seats metric |
| 273 |
DeprecatedVersion: "1.31.0", |
| 274 |
StabilityLevel: compbasemetrics.ALPHA, |
| 275 |
}, |
| 276 |
[]string{priorityLevel, flowSchema}, |
| 277 |
) |
| 278 |
apiserverRequestWaitingSeconds = compbasemetrics.NewHistogramVec( |
| 279 |
&compbasemetrics.HistogramOpts{ |
| 280 |
Namespace: namespace, |
| 281 |
Subsystem: subsystem, |
| 282 |
Name: "request_wait_duration_seconds", |
| 283 |
Help: "Length of time a request spent waiting in its queue", |
| 284 |
Buckets: requestDurationSecondsBuckets, |
| 285 |
StabilityLevel: compbasemetrics.BETA, |
| 286 |
}, |
| 287 |
[]string{priorityLevel, flowSchema, "execute"}, |
| 288 |
) |
| 289 |
apiserverRequestExecutionSeconds = compbasemetrics.NewHistogramVec( |
| 290 |
&compbasemetrics.HistogramOpts{ |
| 291 |
Namespace: namespace, |
| 292 |
Subsystem: subsystem, |
| 293 |
Name: "request_execution_seconds", |
| 294 |
Help: "Duration of initial stage (for a WATCH) or any (for a non-WATCH) stage of request execution in the API Priority and Fairness subsystem", |
| 295 |
Buckets: requestDurationSecondsBuckets, |
| 296 |
StabilityLevel: compbasemetrics.ALPHA, |
| 297 |
}, |
| 298 |
[]string{priorityLevel, flowSchema, "type"}, |
| 299 |
) |
| 300 |
watchCountSamples = compbasemetrics.NewHistogramVec( |
| 301 |
&compbasemetrics.HistogramOpts{ |
| 302 |
Namespace: namespace, |
| 303 |
Subsystem: subsystem, |
| 304 |
Name: "watch_count_samples", |
| 305 |
Help: "count of watchers for mutating requests in API Priority and Fairness", |
| 306 |
Buckets: []float64{0, 1, 10, 100, 1000, 10000}, |
| 307 |
StabilityLevel: compbasemetrics.ALPHA, |
| 308 |
}, |
| 309 |
[]string{priorityLevel, flowSchema}, |
| 310 |
) |
| 311 |
apiserverEpochAdvances = compbasemetrics.NewCounterVec( |
| 312 |
&compbasemetrics.CounterOpts{ |
| 313 |
Namespace: namespace, |
| 314 |
Subsystem: subsystem, |
| 315 |
Name: "epoch_advance_total", |
| 316 |
Help: "Number of times the queueset's progress meter jumped backward", |
| 317 |
StabilityLevel: compbasemetrics.ALPHA, |
| 318 |
}, |
| 319 |
[]string{priorityLevel, "success"}, |
| 320 |
) |
| 321 |
apiserverWorkEstimatedSeats = compbasemetrics.NewHistogramVec( |
| 322 |
&compbasemetrics.HistogramOpts{ |
| 323 |
Namespace: namespace, |
| 324 |
Subsystem: subsystem, |
| 325 |
Name: "work_estimated_seats", |
| 326 |
Help: "Number of estimated seats (maximum of initial and final seats) associated with requests in API Priority and Fairness", |
| 327 |
// the upper bound comes from the maximum number of seats a request |
| 328 |
// can occupy which is currently set at 10. |
| 329 |
Buckets: []float64{1, 2, 4, 10}, |
| 330 |
StabilityLevel: compbasemetrics.ALPHA, |
| 331 |
}, |
| 332 |
[]string{priorityLevel, flowSchema}, |
| 333 |
) |
| 334 |
apiserverDispatchWithNoAccommodation = compbasemetrics.NewCounterVec( |
| 335 |
&compbasemetrics.CounterOpts{ |
| 336 |
Namespace: namespace, |
| 337 |
Subsystem: subsystem, |
| 338 |
Name: "request_dispatch_no_accommodation_total", |
| 339 |
Help: "Number of times a dispatch attempt resulted in a non accommodation due to lack of available seats", |
| 340 |
StabilityLevel: compbasemetrics.ALPHA, |
| 341 |
}, |
| 342 |
[]string{priorityLevel, flowSchema}, |
| 343 |
) |
| 344 |
apiserverNominalConcurrencyLimits = compbasemetrics.NewGaugeVec( |
| 345 |
&compbasemetrics.GaugeOpts{ |
| 346 |
Namespace: namespace, |
| 347 |
Subsystem: subsystem, |
| 348 |
Name: "nominal_limit_seats", |
| 349 |
Help: "Nominal number of execution seats configured for each priority level", |
| 350 |
StabilityLevel: compbasemetrics.BETA, |
| 351 |
}, |
| 352 |
[]string{priorityLevel}, |
| 353 |
) |
| 354 |
apiserverMinimumConcurrencyLimits = compbasemetrics.NewGaugeVec( |
| 355 |
&compbasemetrics.GaugeOpts{ |
| 356 |
Namespace: namespace, |
| 357 |
Subsystem: subsystem, |
| 358 |
Name: "lower_limit_seats", |
| 359 |
Help: "Configured lower bound on number of execution seats available to each priority level", |
| 360 |
StabilityLevel: compbasemetrics.ALPHA, |
| 361 |
}, |
| 362 |
[]string{priorityLevel}, |
| 363 |
) |
| 364 |
apiserverMaximumConcurrencyLimits = compbasemetrics.NewGaugeVec( |
| 365 |
&compbasemetrics.GaugeOpts{ |
| 366 |
Namespace: namespace, |
| 367 |
Subsystem: subsystem, |
| 368 |
Name: "upper_limit_seats", |
| 369 |
Help: "Configured upper bound on number of execution seats available to each priority level", |
| 370 |
StabilityLevel: compbasemetrics.ALPHA, |
| 371 |
}, |
| 372 |
[]string{priorityLevel}, |
| 373 |
) |
| 374 |
ApiserverSeatDemands = NewTimingRatioHistogramVec( |
| 375 |
&compbasemetrics.TimingHistogramOpts{ |
| 376 |
Namespace: namespace, |
| 377 |
Subsystem: subsystem, |
| 378 |
Name: "demand_seats", |
| 379 |
Help: "Observations, at the end of every nanosecond, of (the number of seats each priority level could use) / (nominal number of seats for that level)", |
| 380 |
// Rationale for the bucket boundaries: |
| 381 |
// For 0--1, evenly spaced and not too many; |
| 382 |
// For 1--2, roughly powers of sqrt(sqrt(2)); |
| 383 |
// For 2--6, roughly powers of sqrt(2); |
| 384 |
// We need coverage over 1, but do not want too many buckets. |
| 385 |
Buckets: []float64{0.2, 0.4, 0.6, 0.8, 1, 1.2, 1.4, 1.7, 2, 2.8, 4, 6}, |
| 386 |
StabilityLevel: compbasemetrics.ALPHA, |
| 387 |
}, |
| 388 |
priorityLevel, |
| 389 |
) |
| 390 |
apiserverSeatDemandHighWatermarks = compbasemetrics.NewGaugeVec( |
| 391 |
&compbasemetrics.GaugeOpts{ |
| 392 |
Namespace: namespace, |
| 393 |
Subsystem: subsystem, |
| 394 |
Name: "demand_seats_high_watermark", |
| 395 |
Help: "High watermark, over last adjustment period, of demand_seats", |
| 396 |
StabilityLevel: compbasemetrics.ALPHA, |
| 397 |
}, |
| 398 |
[]string{priorityLevel}, |
| 399 |
) |
| 400 |
apiserverSeatDemandAverages = compbasemetrics.NewGaugeVec( |
| 401 |
&compbasemetrics.GaugeOpts{ |
| 402 |
Namespace: namespace, |
| 403 |
Subsystem: subsystem, |
| 404 |
Name: "demand_seats_average", |
| 405 |
Help: "Time-weighted average, over last adjustment period, of demand_seats", |
| 406 |
StabilityLevel: compbasemetrics.ALPHA, |
| 407 |
}, |
| 408 |
[]string{priorityLevel}, |
| 409 |
) |
| 410 |
apiserverSeatDemandStandardDeviations = compbasemetrics.NewGaugeVec( |
| 411 |
&compbasemetrics.GaugeOpts{ |
| 412 |
Namespace: namespace, |
| 413 |
Subsystem: subsystem, |
| 414 |
Name: "demand_seats_stdev", |
| 415 |
Help: "Time-weighted standard deviation, over last adjustment period, of demand_seats", |
| 416 |
StabilityLevel: compbasemetrics.ALPHA, |
| 417 |
}, |
| 418 |
[]string{priorityLevel}, |
| 419 |
) |
| 420 |
apiserverSeatDemandSmootheds = compbasemetrics.NewGaugeVec( |
| 421 |
&compbasemetrics.GaugeOpts{ |
| 422 |
Namespace: namespace, |
| 423 |
Subsystem: subsystem, |
| 424 |
Name: "demand_seats_smoothed", |
| 425 |
Help: "Smoothed seat demands", |
| 426 |
StabilityLevel: compbasemetrics.ALPHA, |
| 427 |
}, |
| 428 |
[]string{priorityLevel}, |
| 429 |
) |
| 430 |
apiserverSeatDemandTargets = compbasemetrics.NewGaugeVec( |
| 431 |
&compbasemetrics.GaugeOpts{ |
| 432 |
Namespace: namespace, |
| 433 |
Subsystem: subsystem, |
| 434 |
Name: "target_seats", |
| 435 |
Help: "Seat allocation targets", |
| 436 |
StabilityLevel: compbasemetrics.ALPHA, |
| 437 |
}, |
| 438 |
[]string{priorityLevel}, |
| 439 |
) |
| 440 |
apiserverFairFracs = compbasemetrics.NewGauge( |
| 441 |
&compbasemetrics.GaugeOpts{ |
| 442 |
Namespace: namespace, |
| 443 |
Subsystem: subsystem, |
| 444 |
Name: "seat_fair_frac", |
| 445 |
Help: "Fair fraction of server's concurrency to allocate to each priority level that can use it", |
| 446 |
StabilityLevel: compbasemetrics.ALPHA, |
| 447 |
}) |
| 448 |
apiserverCurrentConcurrencyLimits = compbasemetrics.NewGaugeVec( |
| 449 |
&compbasemetrics.GaugeOpts{ |
| 450 |
Namespace: namespace, |
| 451 |
Subsystem: subsystem, |
| 452 |
Name: "current_limit_seats", |
| 453 |
Help: "current derived number of execution seats available to each priority level", |
| 454 |
StabilityLevel: compbasemetrics.ALPHA, |
| 455 |
}, |
| 456 |
[]string{priorityLevel}, |
| 457 |
) |
| 458 |
|
| 459 |
metrics = Registerables{ |
| 460 |
apiserverRejectedRequestsTotal, |
| 461 |
apiserverDispatchedRequestsTotal, |
| 462 |
apiserverCurrentR, |
| 463 |
apiserverDispatchR, |
| 464 |
apiserverLatestS, |
| 465 |
apiserverNextSBounds, |
| 466 |
apiserverNextDiscountedSBounds, |
| 467 |
apiserverCurrentInqueueRequests, |
| 468 |
apiserverCurrentInqueueSeats, |
| 469 |
apiserverRequestQueueLength, |
| 470 |
apiserverRequestConcurrencyLimit, |
| 471 |
apiserverRequestConcurrencyInUse, |
| 472 |
apiserverCurrentExecutingSeats, |
| 473 |
apiserverCurrentExecutingRequests, |
| 474 |
apiserverRequestWaitingSeconds, |
| 475 |
apiserverRequestExecutionSeconds, |
| 476 |
watchCountSamples, |
| 477 |
apiserverEpochAdvances, |
| 478 |
apiserverWorkEstimatedSeats, |
| 479 |
apiserverDispatchWithNoAccommodation, |
| 480 |
apiserverNominalConcurrencyLimits, |
| 481 |
apiserverMinimumConcurrencyLimits, |
| 482 |
apiserverMaximumConcurrencyLimits, |
| 483 |
apiserverSeatDemandHighWatermarks, |
| 484 |
apiserverSeatDemandAverages, |
| 485 |
apiserverSeatDemandStandardDeviations, |
| 486 |
apiserverSeatDemandSmootheds, |
| 487 |
apiserverSeatDemandTargets, |
| 488 |
apiserverFairFracs, |
| 489 |
apiserverCurrentConcurrencyLimits, |
| 490 |
}. |
| 491 |
Append(PriorityLevelExecutionSeatsGaugeVec.metrics()...). |
| 492 |
Append(PriorityLevelConcurrencyGaugeVec.metrics()...). |
| 493 |
Append(readWriteConcurrencyGaugeVec.metrics()...). |
| 494 |
Append(ApiserverSeatDemands.metrics()...) |
| 495 |
) |
| 496 |
|
| 497 |
type indexOnce struct { |
| 498 |
labelValues []string |
| 499 |
once sync.Once |
| 500 |
gauge RatioedGauge |
| 501 |
} |
| 502 |
|
| 503 |
func (io *indexOnce) getGauge() RatioedGauge { |
| 504 |
io.once.Do(func() { |
| 505 |
io.gauge = readWriteConcurrencyGaugeVec.NewForLabelValuesSafe(0, 1, io.labelValues) |
| 506 |
}) |
| 507 |
return io.gauge |
| 508 |
} |
| 509 |
|
| 510 |
var waitingReadonly = indexOnce{labelValues: []string{LabelValueWaiting, epmetrics.ReadOnlyKind}} |
| 511 |
var executingReadonly = indexOnce{labelValues: []string{LabelValueExecuting, epmetrics.ReadOnlyKind}} |
| 512 |
var waitingMutating = indexOnce{labelValues: []string{LabelValueWaiting, epmetrics.MutatingKind}} |
| 513 |
var executingMutating = indexOnce{labelValues: []string{LabelValueExecuting, epmetrics.MutatingKind}} |
| 514 |
|
| 515 |
// GetWaitingReadonlyConcurrency returns the gauge of number of readonly requests waiting / limit on those. |
| 516 |
var GetWaitingReadonlyConcurrency = waitingReadonly.getGauge |
| 517 |
|
| 518 |
// GetExecutingReadonlyConcurrency returns the gauge of number of executing readonly requests / limit on those. |
| 519 |
var GetExecutingReadonlyConcurrency = executingReadonly.getGauge |
| 520 |
|
| 521 |
// GetWaitingMutatingConcurrency returns the gauge of number of mutating requests waiting / limit on those. |
| 522 |
var GetWaitingMutatingConcurrency = waitingMutating.getGauge |
| 523 |
|
| 524 |
// GetExecutingMutatingConcurrency returns the gauge of number of executing mutating requests / limit on those. |
| 525 |
var GetExecutingMutatingConcurrency = executingMutating.getGauge |
| 526 |
|
| 527 |
// AddRequestsInQueues adds the given delta to the gauge of the # of requests in the queues of the specified flowSchema and priorityLevel |
| 528 |
func AddRequestsInQueues(ctx context.Context, priorityLevel, flowSchema string, delta int) { |
| 529 |
apiserverCurrentInqueueRequests.WithLabelValues(priorityLevel, flowSchema).Add(float64(delta)) |
| 530 |
} |
| 531 |
|
| 532 |
// AddSeatsInQueues adds the given delta to the gauge of the # of seats in the queues of the specified flowSchema and priorityLevel |
| 533 |
func AddSeatsInQueues(ctx context.Context, priorityLevel, flowSchema string, delta int) { |
| 534 |
apiserverCurrentInqueueSeats.WithLabelValues(priorityLevel, flowSchema).Add(float64(delta)) |
| 535 |
} |
| 536 |
|
| 537 |
// AddRequestsExecuting adds the given delta to the gauge of executing requests of the given flowSchema and priorityLevel |
| 538 |
func AddRequestsExecuting(ctx context.Context, priorityLevel, flowSchema string, delta int) { |
| 539 |
apiserverCurrentExecutingRequests.WithLabelValues(priorityLevel, flowSchema).Add(float64(delta)) |
| 540 |
} |
| 541 |
|
| 542 |
// SetCurrentR sets the current-R (virtualTime) gauge for the given priority level |
| 543 |
func SetCurrentR(priorityLevel string, r float64) { |
| 544 |
apiserverCurrentR.WithLabelValues(priorityLevel).Set(r) |
| 545 |
} |
| 546 |
|
| 547 |
// SetLatestS sets the latest-S (virtual time of dispatched request) gauge for the given priority level |
| 548 |
func SetDispatchMetrics(priorityLevel string, r, s, sMin, sMax, discountedSMin, discountedSMax float64) { |
| 549 |
apiserverDispatchR.WithLabelValues(priorityLevel).Set(r) |
| 550 |
apiserverLatestS.WithLabelValues(priorityLevel).Set(s) |
| 551 |
apiserverNextSBounds.WithLabelValues(priorityLevel, "min").Set(sMin) |
| 552 |
apiserverNextSBounds.WithLabelValues(priorityLevel, "max").Set(sMax) |
| 553 |
apiserverNextDiscountedSBounds.WithLabelValues(priorityLevel, "min").Set(discountedSMin) |
| 554 |
apiserverNextDiscountedSBounds.WithLabelValues(priorityLevel, "max").Set(discountedSMax) |
| 555 |
} |
| 556 |
|
| 557 |
// AddSeatConcurrencyInUse adds the given delta to the gauge of seats in use by |
| 558 |
// the currently executing requests of the given flowSchema and priorityLevel |
| 559 |
func AddSeatConcurrencyInUse(priorityLevel, flowSchema string, delta int) { |
| 560 |
apiserverCurrentExecutingSeats.WithLabelValues(priorityLevel, flowSchema).Add(float64(delta)) |
| 561 |
apiserverRequestConcurrencyInUse.WithLabelValues(priorityLevel, flowSchema).Add(float64(delta)) |
| 562 |
} |
| 563 |
|
| 564 |
// AddReject increments the # of rejected requests for flow control |
| 565 |
func AddReject(ctx context.Context, priorityLevel, flowSchema, reason string) { |
| 566 |
apiserverRejectedRequestsTotal.WithContext(ctx).WithLabelValues(priorityLevel, flowSchema, reason).Add(1) |
| 567 |
} |
| 568 |
|
| 569 |
// AddDispatch increments the # of dispatched requests for flow control |
| 570 |
func AddDispatch(ctx context.Context, priorityLevel, flowSchema string) { |
| 571 |
apiserverDispatchedRequestsTotal.WithContext(ctx).WithLabelValues(priorityLevel, flowSchema).Add(1) |
| 572 |
} |
| 573 |
|
| 574 |
// ObserveQueueLength observes the queue length for flow control |
| 575 |
func ObserveQueueLength(ctx context.Context, priorityLevel, flowSchema string, length int) { |
| 576 |
apiserverRequestQueueLength.WithContext(ctx).WithLabelValues(priorityLevel, flowSchema).Observe(float64(length)) |
| 577 |
} |
| 578 |
|
| 579 |
// ObserveWaitingDuration observes the queue length for flow control |
| 580 |
func ObserveWaitingDuration(ctx context.Context, priorityLevel, flowSchema, execute string, waitTime time.Duration) { |
| 581 |
apiserverRequestWaitingSeconds.WithContext(ctx).WithLabelValues(priorityLevel, flowSchema, execute).Observe(waitTime.Seconds()) |
| 582 |
} |
| 583 |
|
| 584 |
// ObserveExecutionDuration observes the execution duration for flow control |
| 585 |
func ObserveExecutionDuration(ctx context.Context, priorityLevel, flowSchema string, executionTime time.Duration) { |
| 586 |
reqType := "regular" |
| 587 |
if requestInfo, ok := apirequest.RequestInfoFrom(ctx); ok && requestInfo.Verb == "watch" { |
| 588 |
reqType = requestInfo.Verb |
| 589 |
} |
| 590 |
apiserverRequestExecutionSeconds.WithContext(ctx).WithLabelValues(priorityLevel, flowSchema, reqType).Observe(executionTime.Seconds()) |
| 591 |
} |
| 592 |
|
| 593 |
// ObserveWatchCount notes a sampling of a watch count |
| 594 |
func ObserveWatchCount(ctx context.Context, priorityLevel, flowSchema string, count int) { |
| 595 |
watchCountSamples.WithLabelValues(priorityLevel, flowSchema).Observe(float64(count)) |
| 596 |
} |
| 597 |
|
| 598 |
// AddEpochAdvance notes an advance of the progress meter baseline for a given priority level |
| 599 |
func AddEpochAdvance(ctx context.Context, priorityLevel string, success bool) { |
| 600 |
apiserverEpochAdvances.WithContext(ctx).WithLabelValues(priorityLevel, strconv.FormatBool(success)).Inc() |
| 601 |
} |
| 602 |
|
| 603 |
// ObserveWorkEstimatedSeats notes a sampling of estimated seats associated with a request |
| 604 |
func ObserveWorkEstimatedSeats(priorityLevel, flowSchema string, seats int) { |
| 605 |
apiserverWorkEstimatedSeats.WithLabelValues(priorityLevel, flowSchema).Observe(float64(seats)) |
| 606 |
} |
| 607 |
|
| 608 |
// AddDispatchWithNoAccommodation keeps track of number of times dispatch attempt results |
| 609 |
// in a non accommodation due to lack of available seats. |
| 610 |
func AddDispatchWithNoAccommodation(priorityLevel, flowSchema string) { |
| 611 |
apiserverDispatchWithNoAccommodation.WithLabelValues(priorityLevel, flowSchema).Inc() |
| 612 |
} |
| 613 |
|
| 614 |
func SetPriorityLevelConfiguration(priorityLevel string, nominalCL, minCL, maxCL int) { |
| 615 |
apiserverRequestConcurrencyLimit.WithLabelValues(priorityLevel).Set(float64(nominalCL)) |
| 616 |
apiserverNominalConcurrencyLimits.WithLabelValues(priorityLevel).Set(float64(nominalCL)) |
| 617 |
apiserverMinimumConcurrencyLimits.WithLabelValues(priorityLevel).Set(float64(minCL)) |
| 618 |
apiserverMaximumConcurrencyLimits.WithLabelValues(priorityLevel).Set(float64(maxCL)) |
| 619 |
} |
| 620 |
|
| 621 |
func NotePriorityLevelConcurrencyAdjustment(priorityLevel string, seatDemandHWM, seatDemandAvg, seatDemandStdev, seatDemandSmoothed, seatDemandTarget float64, currentCL int) { |
| 622 |
apiserverSeatDemandHighWatermarks.WithLabelValues(priorityLevel).Set(seatDemandHWM) |
| 623 |
apiserverSeatDemandAverages.WithLabelValues(priorityLevel).Set(seatDemandAvg) |
| 624 |
apiserverSeatDemandStandardDeviations.WithLabelValues(priorityLevel).Set(seatDemandStdev) |
| 625 |
apiserverSeatDemandSmootheds.WithLabelValues(priorityLevel).Set(seatDemandSmoothed) |
| 626 |
apiserverSeatDemandTargets.WithLabelValues(priorityLevel).Set(seatDemandTarget) |
| 627 |
apiserverCurrentConcurrencyLimits.WithLabelValues(priorityLevel).Set(float64(currentCL)) |
| 628 |
} |
| 629 |
|
| 630 |
func SetFairFrac(fairFrac float64) { |
| 631 |
apiserverFairFracs.Set(fairFrac) |
| 632 |
} |