-
Notifications
You must be signed in to change notification settings - Fork 12
/
server_metrics.go
312 lines (284 loc) · 12.8 KB
/
server_metrics.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
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
// Copyright 2022 The searKing Author. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package otelgrpc
import (
"context"
"time"
errors_ "github.com/searKing/golang/go/errors"
slices_ "github.com/searKing/golang/go/exp/slices"
net_ "github.com/searKing/golang/go/net"
otelgrpc_ "go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc"
"go.opentelemetry.io/otel/attribute"
"go.opentelemetry.io/otel/metric"
"google.golang.org/grpc"
)
// ServerMetrics represents a collection of metrics to be registered on a
// Prometheus metrics registry for a gRPC server.
type ServerMetrics struct {
ServerHostport string
serverStartedCounter metric.Int64Counter
serverHandledCounter metric.Int64Counter
serverStreamMsgReceived metric.Int64Counter
serverStreamMsgSent metric.Int64Counter
serverHandledTimeHistogramEnabled bool
serverHandledTimeHistogram metric.Float64Histogram
serverStreamReceiveTimeHistogramEnabled bool
serverStreamReceiveTimeHistogram metric.Float64Histogram
serverStreamReceiveSizeHistogramEnabled bool
serverStreamReceiveSizeHistogram metric.Int64Histogram
serverStreamSendTimeHistogramEnabled bool
serverStreamSendTimeHistogram metric.Float64Histogram
serverStreamSendSizeHistogramEnabled bool
serverStreamSendSizeHistogram metric.Int64Histogram
}
// NewServerMetrics returns a ServerMetrics object. Use a new instance of
// ServerMetrics when not using the default Prometheus metrics registry, for
// example when wanting to control which metrics are added to a registry as
// opposed to automatically adding metrics via init functions.
func NewServerMetrics(opts ...metric.InstrumentOption) *ServerMetrics {
m := &ServerMetrics{}
errors_.Must(m.ResetCounter(opts...))
return m
}
// ResetCounter recreate recording of all counters of RPCs.
func (m *ServerMetrics) ResetCounter(opts ...metric.InstrumentOption) (err error) {
int64Opts := slices_.MapFunc(opts, func(e metric.InstrumentOption) metric.Int64CounterOption { return e })
var addr string
if ip, err := net_.ListenIP(); err == nil {
addr = ip.String()
}
m.ServerHostport = addr
// "grpc_type", "grpc_service", "grpc_method"
m.serverStartedCounter, err = Meter().Int64Counter(
"grpc_server_started_total",
func() []metric.Int64CounterOption {
var options []metric.Int64CounterOption
options = append(options, metric.WithDescription("Total number of RPCs started on the server."))
options = append(options, int64Opts...)
return options
}()...)
if err != nil {
return err
}
// "grpc_type", "grpc_service", "grpc_method", "grpc_code"
m.serverHandledCounter, err = Meter().Int64Counter(
"grpc_server_handled_total",
func() []metric.Int64CounterOption {
var options []metric.Int64CounterOption
options = append(options, metric.WithDescription("Total number of RPCs completed by the server, regardless of success or failure."))
options = append(options, int64Opts...)
return options
}()...)
if err != nil {
return err
}
// "grpc_type", "grpc_service", "grpc_method"
m.serverStreamMsgReceived, err = Meter().Int64Counter(
"grpc_server_msg_received_total",
func() []metric.Int64CounterOption {
var options []metric.Int64CounterOption
options = append(options, metric.WithDescription("Total number of RPC stream messages received by the server."))
options = append(options, int64Opts...)
return options
}()...)
if err != nil {
return err
}
// "grpc_type", "grpc_service", "grpc_method"
m.serverStreamMsgSent, err = Meter().Int64Counter(
"grpc_server_msg_sent_total",
func() []metric.Int64CounterOption {
var options []metric.Int64CounterOption
options = append(options, metric.WithDescription("Total number of gRPC stream messages sent by the server."))
options = append(options, int64Opts...)
return options
}()...)
return err
}
// EnableServerHandledTimeHistogram turns on recording of handling time of RPCs.
// Histogram metrics can be very expensive for Prometheus to retain and query.
func (m *ServerMetrics) EnableServerHandledTimeHistogram(opts ...metric.Float64HistogramOption) (err error) {
var options []metric.Float64HistogramOption
options = append(options,
metric.WithDescription("Histogram of response latency (seconds) of the gRPC until it is finished by the server."),
metric.WithUnit("s"))
options = append(options, opts...)
if !m.serverHandledTimeHistogramEnabled {
// https://github.com/open-telemetry/opentelemetry-go/issues/1280
m.serverHandledTimeHistogram, err = Meter().Float64Histogram("grpc_server_handling_seconds", options...)
if err != nil {
return err
}
}
m.serverHandledTimeHistogramEnabled = true
return nil
}
// EnableServerStreamReceiveTimeHistogram turns on recording of single message receive time of streaming RPCs.
// Histogram metrics can be very expensive for Prometheus to retain and query.
func (m *ServerMetrics) EnableServerStreamReceiveTimeHistogram(opts ...metric.Float64HistogramOption) (err error) {
var options []metric.Float64HistogramOption
options = append(options,
metric.WithDescription("Histogram of response latency (seconds) of the gRPC single message receive."),
metric.WithUnit("s"))
options = append(options, opts...)
if !m.serverStreamReceiveTimeHistogramEnabled {
// https://github.com/open-telemetry/opentelemetry-go/issues/1280
m.serverStreamReceiveTimeHistogram, err = Meter().Float64Histogram("grpc_server_msg_recv_handling_seconds", options...)
if err != nil {
return err
}
}
m.serverStreamReceiveTimeHistogramEnabled = true
return nil
}
// EnableServerStreamReceiveSizeHistogram turns on recording of single message receive size of streaming RPCs.
// Histogram metrics can be very expensive for Prometheus to retain and query.
func (m *ServerMetrics) EnableServerStreamReceiveSizeHistogram(opts ...metric.Int64HistogramOption) (err error) {
var options []metric.Int64HistogramOption
options = append(options,
metric.WithDescription("Histogram of message size (bytes) of the gRPC single message receive."),
metric.WithUnit(uBytes))
options = append(options, opts...)
if !m.serverStreamReceiveSizeHistogramEnabled {
// https://github.com/open-telemetry/opentelemetry-go/issues/1280
m.serverStreamReceiveSizeHistogram, err = Meter().Int64Histogram("grpc_server_msg_recv_handling_bytes", options...)
if err != nil {
return err
}
}
m.serverStreamReceiveSizeHistogramEnabled = true
return nil
}
// EnableServerStreamSendTimeHistogram turns on recording of single message send time of streaming RPCs.
// Histogram metrics can be very expensive for Prometheus to retain and query.
func (m *ServerMetrics) EnableServerStreamSendTimeHistogram(opts ...metric.Float64HistogramOption) (err error) {
var options []metric.Float64HistogramOption
options = append(options,
metric.WithDescription("Histogram of response latency (seconds) of the gRPC single message send."),
metric.WithUnit("s"))
options = append(options, opts...)
if !m.serverStreamSendTimeHistogramEnabled {
// https://github.com/open-telemetry/opentelemetry-go/issues/1280
m.serverStreamSendTimeHistogram, err = Meter().Float64Histogram("grpc_server_msg_send_handling_seconds", options...)
if err != nil {
return err
}
}
m.serverStreamSendTimeHistogramEnabled = true
return nil
}
// EnableServerStreamSendSizeHistogram turns on recording of single message send size of streaming RPCs.
// Histogram metrics can be very expensive for Prometheus to retain and query.
func (m *ServerMetrics) EnableServerStreamSendSizeHistogram(opts ...metric.Int64HistogramOption) (err error) {
var options []metric.Int64HistogramOption
options = append(options,
metric.WithDescription("Histogram of message size (bytes) of the gRPC single message send."),
metric.WithUnit(uBytes))
options = append(options, opts...)
if !m.serverStreamSendSizeHistogramEnabled {
// https://github.com/open-telemetry/opentelemetry-go/issues/1280
m.serverStreamSendSizeHistogram, err = Meter().Int64Histogram("grpc_server_msg_send_handling_bytes", options...)
if err != nil {
return err
}
}
m.serverStreamSendSizeHistogramEnabled = true
return nil
}
// UnaryServerInterceptor is a gRPC server-side interceptor that provides Prometheus monitoring for Unary RPCs.
func (m *ServerMetrics) UnaryServerInterceptor() func(ctx context.Context, req any, info *grpc.UnaryServerInfo, handler grpc.UnaryHandler) (any, error) {
return func(ctx context.Context, req any, info *grpc.UnaryServerInfo, handler grpc.UnaryHandler) (any, error) {
monitor := newServerReporter(ctx, m, Unary, info.FullMethod, peerFromCtx(ctx), m.ServerHostport)
monitor.ReceivedMessage(ctx, req)
resp, err := handler(ctx, req)
st, _ := FromError(err)
monitor.Handled(ctx, st.Code())
if err == nil {
monitor.SentMessage(ctx, resp)
}
return resp, err
}
}
// StreamServerInterceptor is a gRPC server-side interceptor that provides Prometheus monitoring for Streaming RPCs.
func (m *ServerMetrics) StreamServerInterceptor() func(srv any, ss grpc.ServerStream, info *grpc.StreamServerInfo, handler grpc.StreamHandler) error {
return func(srv any, ss grpc.ServerStream, info *grpc.StreamServerInfo, handler grpc.StreamHandler) error {
monitor := newServerReporter(ss.Context(), m, streamRPCType(info), info.FullMethod, peerFromCtx(ss.Context()), m.ServerHostport)
err := handler(srv, &monitoredServerStream{ss, monitor})
st, _ := FromError(err)
monitor.Handled(ss.Context(), st.Code())
return err
}
}
// InitializeMetrics initializes all metrics, with their appropriate null
// value, for all gRPC methods registered on a gRPC server. This is useful, to
// ensure that all metrics exist when collecting and querying.
func (m *ServerMetrics) InitializeMetrics(ctx context.Context, server *grpc.Server) {
serviceInfo := server.GetServiceInfo()
for serviceName, info := range serviceInfo {
for _, mInfo := range info.Methods {
preRegisterMethod(ctx, m, serviceName, &mInfo)
}
}
}
func streamRPCType(info *grpc.StreamServerInfo) grpcType {
if info.IsServerStream && !info.IsClientStream {
return ServerStream
} else if info.IsClientStream && !info.IsServerStream {
return ClientStream
}
return BidiStream
}
// monitoredStream wraps grpc.ServerStream allowing each Sent/Recv of message to increment counters.
type monitoredServerStream struct {
grpc.ServerStream
monitor *serverReporter
}
func (s *monitoredServerStream) SendMsg(m any) error {
now := time.Now()
err := s.ServerStream.SendMsg(m)
s.monitor.SendMessageTimer(context.Background(), now)
if err == nil {
s.monitor.SentMessage(context.Background(), m)
}
return err
}
func (s *monitoredServerStream) RecvMsg(m any) error {
now := time.Now()
err := s.ServerStream.RecvMsg(m)
s.monitor.ReceiveMessageTimer(context.Background(), now)
if err == nil {
s.monitor.ReceivedMessage(context.Background(), m)
}
return err
}
// preRegisterMethod is invoked on Register of a Server, allowing all gRPC services labels to be pre-populated.
func preRegisterMethod(ctx context.Context, metrics *ServerMetrics, serviceName string, mInfo *grpc.MethodInfo) {
filter := AttrsFilter
if filter == nil {
filter = func(attrs ...attribute.KeyValue) []attribute.KeyValue { return attrs }
}
// These are just references (no increments), as just referencing will create the labels but not set values.
_, attrs := spanInfo(mInfo.Name, ":0", metrics.ServerHostport, typeFromMethodInfo(mInfo), false)
metrics.serverStartedCounter.Add(ctx, 0, metric.WithAttributes(filter(attrs...)...))
metrics.serverStreamMsgReceived.Add(ctx, 0, metric.WithAttributes(filter(append(attrs, otelgrpc_.RPCMessageTypeReceived)...)...))
metrics.serverStreamMsgSent.Add(ctx, 0, metric.WithAttributes(filter(append(attrs, otelgrpc_.RPCMessageTypeSent)...)...))
for _, code := range allCodes {
metrics.serverHandledCounter.Add(ctx, 0, metric.WithAttributes(filter(append(attrs, statusCodeAttr(code))...)...))
if metrics.serverHandledTimeHistogramEnabled {
metrics.serverHandledTimeHistogram.Record(ctx, -1, metric.WithAttributes(filter(append(attrs, statusCodeAttr(code))...)...))
}
}
if metrics.serverStreamReceiveTimeHistogramEnabled {
metrics.serverStreamReceiveTimeHistogram.Record(ctx, -1, metric.WithAttributes(filter(append(attrs, otelgrpc_.RPCMessageTypeReceived)...)...))
}
if metrics.serverStreamReceiveSizeHistogramEnabled {
metrics.serverStreamReceiveSizeHistogram.Record(ctx, -1, metric.WithAttributes(filter(append(attrs, otelgrpc_.RPCMessageTypeReceived)...)...))
}
if metrics.serverStreamSendTimeHistogramEnabled {
metrics.serverStreamSendTimeHistogram.Record(ctx, -1, metric.WithAttributes(filter(append(attrs, otelgrpc_.RPCMessageTypeSent)...)...))
}
if metrics.serverStreamSendSizeHistogramEnabled {
metrics.serverStreamSendSizeHistogram.Record(ctx, -1, metric.WithAttributes(filter(append(attrs, otelgrpc_.RPCMessageTypeSent)...)...))
}
}