/
health_request_rate_limiter.go
204 lines (175 loc) · 7.22 KB
/
health_request_rate_limiter.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
// The MIT License
//
// Copyright (c) 2020 Temporal Technologies Inc. All rights reserved.
//
// Copyright (c) 2020 Uber Technologies, Inc.
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in
// all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
// THE SOFTWARE.
package client
import (
"context"
"encoding/json"
"math"
"sync/atomic"
"time"
"go.temporal.io/server/common/log"
"go.temporal.io/server/common/log/tag"
"go.temporal.io/server/common/metrics"
"go.temporal.io/server/common/persistence"
"go.temporal.io/server/common/primitives/timestamp"
"go.temporal.io/server/common/quotas"
)
const (
DefaultRefreshInterval = 10 * time.Second
DefaultRateBurstRatio = 1.0
DefaultInitialRateMultiplier = 1.0
)
type (
HealthRequestRateLimiterImpl struct {
enabled atomic.Bool
params DynamicRateLimitingParams // dynamic config map
curOptions dynamicRateLimitingOptions // current dynamic config values (updated on refresh)
rateLimiter *quotas.RateLimiterImpl
healthSignals persistence.HealthSignalAggregator
refreshTimer *time.Ticker
rateFn quotas.RateFn
rateToBurstRatio float64
curRateMultiplier float64
metricsHandler metrics.Handler
logger log.Logger
}
dynamicRateLimitingOptions struct {
Enabled bool
RefreshInterval string // string returned by json.Unmarshal will be parsed into a duration
// thresholds which should trigger backoff if exceeded
LatencyThreshold float64
ErrorThreshold float64
// if either threshold is exceeded, the current rate multiplier will be reduced by this amount
RateBackoffStepSize float64
// when the system is healthy and current rate < max rate, the current rate multiplier will be
// increased by this amount
RateIncreaseStepSize float64
RateMultiMax float64
RateMultiMin float64
}
)
var _ quotas.RequestRateLimiter = (*HealthRequestRateLimiterImpl)(nil)
func NewHealthRequestRateLimiterImpl(
healthSignals persistence.HealthSignalAggregator,
rateFn quotas.RateFn,
params DynamicRateLimitingParams,
metricsHandler metrics.Handler,
logger log.Logger,
) *HealthRequestRateLimiterImpl {
limiter := &HealthRequestRateLimiterImpl{
enabled: atomic.Bool{},
rateLimiter: quotas.NewRateLimiter(rateFn(), int(DefaultRateBurstRatio*rateFn())),
healthSignals: healthSignals,
rateFn: rateFn,
params: params,
refreshTimer: time.NewTicker(DefaultRefreshInterval),
rateToBurstRatio: DefaultRateBurstRatio,
curRateMultiplier: DefaultInitialRateMultiplier,
metricsHandler: metricsHandler,
logger: logger,
}
limiter.refreshDynamicParams()
return limiter
}
func (rl *HealthRequestRateLimiterImpl) Allow(now time.Time, request quotas.Request) bool {
rl.maybeRefresh()
if !rl.enabled.Load() {
return true
}
return rl.rateLimiter.AllowN(now, request.Token)
}
func (rl *HealthRequestRateLimiterImpl) Reserve(now time.Time, request quotas.Request) quotas.Reservation {
rl.maybeRefresh()
if !rl.enabled.Load() {
return quotas.NoopReservation
}
return rl.rateLimiter.ReserveN(now, request.Token)
}
func (rl *HealthRequestRateLimiterImpl) Wait(ctx context.Context, request quotas.Request) error {
rl.maybeRefresh()
if !rl.enabled.Load() {
return nil
}
return rl.rateLimiter.WaitN(ctx, request.Token)
}
func (rl *HealthRequestRateLimiterImpl) maybeRefresh() {
select {
case <-rl.refreshTimer.C:
rl.refreshDynamicParams()
if rl.enabled.Load() {
rl.refreshRate()
}
rl.updateRefreshTimer()
default:
// no-op
}
}
func (rl *HealthRequestRateLimiterImpl) refreshRate() {
if rl.latencyThresholdExceeded() || rl.errorThresholdExceeded() {
// limit exceeded, do backoff
rl.curRateMultiplier = math.Max(rl.curOptions.RateMultiMin, rl.curRateMultiplier-rl.curOptions.RateBackoffStepSize)
rl.metricsHandler.Gauge(metrics.DynamicRateLimiterMultiplier.Name()).Record(rl.curRateMultiplier)
rl.logger.Info("Health threshold exceeded, reducing rate limit.", tag.NewFloat64("newMulti", rl.curRateMultiplier), tag.NewFloat64("newRate", rl.rateLimiter.Rate()), tag.NewFloat64("latencyAvg", rl.healthSignals.AverageLatency()), tag.NewFloat64("errorRatio", rl.healthSignals.ErrorRatio()))
} else if rl.curRateMultiplier < rl.curOptions.RateMultiMax {
// already doing backoff and under thresholds, increase limit
rl.curRateMultiplier = math.Min(rl.curOptions.RateMultiMax, rl.curRateMultiplier+rl.curOptions.RateIncreaseStepSize)
rl.metricsHandler.Gauge(metrics.DynamicRateLimiterMultiplier.Name()).Record(rl.curRateMultiplier)
rl.logger.Info("System healthy, increasing rate limit.", tag.NewFloat64("newMulti", rl.curRateMultiplier), tag.NewFloat64("newRate", rl.rateLimiter.Rate()), tag.NewFloat64("latencyAvg", rl.healthSignals.AverageLatency()), tag.NewFloat64("errorRatio", rl.healthSignals.ErrorRatio()))
}
// Always set rate to pickup changes to underlying rate limit dynamic config
rl.rateLimiter.SetRPS(rl.curRateMultiplier * rl.rateFn())
rl.rateLimiter.SetBurst(int(rl.rateToBurstRatio * rl.rateFn()))
}
func (rl *HealthRequestRateLimiterImpl) refreshDynamicParams() {
var options dynamicRateLimitingOptions
b, err := json.Marshal(rl.params())
if err != nil {
rl.logger.Warn("Error marshalling dynamic rate limiting params. Dynamic rate limiting is disabled.", tag.Error(err))
rl.enabled.Store(false)
return
}
err = json.Unmarshal(b, &options)
if err != nil {
rl.logger.Warn("Error unmarshalling dynamic rate limiting params. Dynamic rate limiting is disabled.", tag.Error(err))
rl.enabled.Store(false)
return
}
rl.enabled.Store(options.Enabled)
rl.curOptions = options
}
func (rl *HealthRequestRateLimiterImpl) updateRefreshTimer() {
if len(rl.curOptions.RefreshInterval) > 0 {
if refreshDuration, err := timestamp.ParseDurationDefaultSeconds(rl.curOptions.RefreshInterval); err != nil {
rl.logger.Warn("Error parsing dynamic rate limit refreshInterval timestamp. Using previous value.", tag.Error(err))
} else {
rl.refreshTimer.Reset(refreshDuration)
}
}
}
func (rl *HealthRequestRateLimiterImpl) latencyThresholdExceeded() bool {
return rl.curOptions.LatencyThreshold > 0 && rl.healthSignals.AverageLatency() > rl.curOptions.LatencyThreshold
}
func (rl *HealthRequestRateLimiterImpl) errorThresholdExceeded() bool {
return rl.curOptions.ErrorThreshold > 0 && rl.healthSignals.ErrorRatio() > rl.curOptions.ErrorThreshold
}