roles.go 11.2 KB
Newer Older
1 2 3 4 5 6
// Licensed to the LF AI & Data foundation under one
// or more contributor license agreements. See the NOTICE file
// distributed with this work for additional information
// regarding copyright ownership. The ASF licenses this file
// to you under the Apache License, Version 2.0 (the
// "License"); you may not use this file except in compliance
7 8
// with the License. You may obtain a copy of the License at
//
9
//     http://www.apache.org/licenses/LICENSE-2.0
10
//
11 12 13 14 15
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
16 17 18 19 20 21

package roles

import (
	"context"
	"fmt"
D
dragondriver 已提交
22
	"net/http"
23 24 25 26 27 28 29
	"os"
	"os/signal"
	"path"
	"strings"
	"sync"
	"syscall"

D
dragondriver 已提交
30 31
	"github.com/milvus-io/milvus/internal/util/healthz"

32
	"github.com/milvus-io/milvus/internal/util/metricsinfo"
S
sunby 已提交
33
	"go.uber.org/zap"
34

35
	"github.com/milvus-io/milvus/cmd/components"
36
	"github.com/milvus-io/milvus/internal/datacoord"
37
	"github.com/milvus-io/milvus/internal/datanode"
38
	"github.com/milvus-io/milvus/internal/indexcoord"
39 40 41 42 43
	"github.com/milvus-io/milvus/internal/indexnode"
	"github.com/milvus-io/milvus/internal/log"
	"github.com/milvus-io/milvus/internal/logutil"
	"github.com/milvus-io/milvus/internal/metrics"
	"github.com/milvus-io/milvus/internal/msgstream"
C
Cai Yudong 已提交
44
	"github.com/milvus-io/milvus/internal/proxy"
45
	"github.com/milvus-io/milvus/internal/querycoord"
46
	"github.com/milvus-io/milvus/internal/querynode"
47
	"github.com/milvus-io/milvus/internal/rootcoord"
48 49 50 51
	"github.com/milvus-io/milvus/internal/util/paramtable"
	"github.com/milvus-io/milvus/internal/util/trace"
)

Y
yukun 已提交
52
func newMsgFactory(localMsg bool) msgstream.Factory {
53
	if localMsg {
Y
yukun 已提交
54
		return msgstream.NewRmsFactory()
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
	}
	return msgstream.NewPmsFactory()
}

type MilvusRoles struct {
	EnableRootCoord      bool `env:"ENABLE_ROOT_COORD"`
	EnableProxy          bool `env:"ENABLE_PROXY"`
	EnableQueryCoord     bool `env:"ENABLE_QUERY_COORD"`
	EnableQueryNode      bool `env:"ENABLE_QUERY_NODE"`
	EnableDataCoord      bool `env:"ENABLE_DATA_COORD"`
	EnableDataNode       bool `env:"ENABLE_DATA_NODE"`
	EnableIndexCoord     bool `env:"ENABLE_INDEX_COORD"`
	EnableIndexNode      bool `env:"ENABLE_INDEX_NODE"`
	EnableMsgStreamCoord bool `env:"ENABLE_MSGSTREAM_COORD"`
}

func (mr *MilvusRoles) EnvValue(env string) bool {
	env = strings.ToLower(env)
	env = strings.Trim(env, " ")
	return env == "1" || env == "true"
}

func (mr *MilvusRoles) setLogConfigFilename(filename string) *log.Config {
	paramtable.Params.Init()
	cfg := paramtable.Params.LogConfig
	if len(cfg.File.RootPath) != 0 {
		cfg.File.Filename = path.Join(cfg.File.RootPath, filename)
	} else {
		cfg.File.Filename = ""
	}
	return cfg
}

func (mr *MilvusRoles) runRootCoord(ctx context.Context, localMsg bool) *components.RootCoord {
	var rc *components.RootCoord
	var wg sync.WaitGroup

	wg.Add(1)
	go func() {
94
		rootcoord.Params.Init()
95

X
Xiaofan 已提交
96 97
		f := setLoggerFunc(localMsg)
		rootcoord.Params.SetLogConfig(f)
Y
yukun 已提交
98
		factory := newMsgFactory(localMsg)
99 100 101 102 103 104 105
		var err error
		rc, err = components.NewRootCoord(ctx, factory)
		if err != nil {
			panic(err)
		}
		wg.Done()
		_ = rc.Run()
D
dragondriver 已提交
106 107 108
		if !localMsg {
			http.Handle(healthz.HealthzRouterPath, &componentsHealthzHandler{component: rc})
		}
109 110 111 112 113 114 115
	}()
	wg.Wait()

	metrics.RegisterRootCoord()
	return rc
}

X
XuanYang-cn 已提交
116
func (mr *MilvusRoles) runProxy(ctx context.Context, localMsg bool, alias string) *components.Proxy {
117 118 119 120 121
	var pn *components.Proxy
	var wg sync.WaitGroup

	wg.Add(1)
	go func() {
C
Cai Yudong 已提交
122 123
		proxy.Params.InitAlias(alias)
		proxy.Params.Init()
124

X
Xiaofan 已提交
125 126
		f := setLoggerFunc(localMsg)
		proxy.Params.SetLogConfig(f)
Y
yukun 已提交
127
		factory := newMsgFactory(localMsg)
128 129 130 131 132 133 134
		var err error
		pn, err = components.NewProxy(ctx, factory)
		if err != nil {
			panic(err)
		}
		wg.Done()
		_ = pn.Run()
D
dragondriver 已提交
135 136 137
		if !localMsg {
			http.Handle(healthz.HealthzRouterPath, &componentsHealthzHandler{component: pn})
		}
138 139 140
	}()
	wg.Wait()

D
dragondriver 已提交
141
	metrics.RegisterProxy()
142 143 144 145 146 147 148 149 150
	return pn
}

func (mr *MilvusRoles) runQueryCoord(ctx context.Context, localMsg bool) *components.QueryCoord {
	var qs *components.QueryCoord
	var wg sync.WaitGroup

	wg.Add(1)
	go func() {
151
		querycoord.Params.Init()
152

X
Xiaofan 已提交
153 154
		f := setLoggerFunc(localMsg)
		querycoord.Params.SetLogConfig(f)
Y
yukun 已提交
155
		factory := newMsgFactory(localMsg)
156 157 158 159 160 161 162
		var err error
		qs, err = components.NewQueryCoord(ctx, factory)
		if err != nil {
			panic(err)
		}
		wg.Done()
		_ = qs.Run()
D
dragondriver 已提交
163 164 165
		if !localMsg {
			http.Handle(healthz.HealthzRouterPath, &componentsHealthzHandler{component: qs})
		}
166 167 168 169 170 171 172
	}()
	wg.Wait()

	metrics.RegisterQueryCoord()
	return qs
}

X
XuanYang-cn 已提交
173
func (mr *MilvusRoles) runQueryNode(ctx context.Context, localMsg bool, alias string) *components.QueryNode {
174 175 176 177 178
	var qn *components.QueryNode
	var wg sync.WaitGroup

	wg.Add(1)
	go func() {
X
XuanYang-cn 已提交
179
		querynode.Params.InitAlias(alias)
180 181
		querynode.Params.Init()

X
Xiaofan 已提交
182 183
		f := setLoggerFunc(localMsg)
		querynode.Params.SetLogConfig(f)
Y
yukun 已提交
184
		factory := newMsgFactory(localMsg)
185 186 187 188 189 190 191
		var err error
		qn, err = components.NewQueryNode(ctx, factory)
		if err != nil {
			panic(err)
		}
		wg.Done()
		_ = qn.Run()
D
dragondriver 已提交
192 193 194
		if !localMsg {
			http.Handle(healthz.HealthzRouterPath, &componentsHealthzHandler{component: qn})
		}
195 196 197 198 199 200 201 202 203 204 205 206 207
	}()
	wg.Wait()

	metrics.RegisterQueryNode()
	return qn
}

func (mr *MilvusRoles) runDataCoord(ctx context.Context, localMsg bool) *components.DataCoord {
	var ds *components.DataCoord
	var wg sync.WaitGroup

	wg.Add(1)
	go func() {
208
		datacoord.Params.Init()
209

X
Xiaofan 已提交
210 211
		f := setLoggerFunc(localMsg)
		datacoord.Params.SetLogConfig(f)
Y
yukun 已提交
212
		factory := newMsgFactory(localMsg)
213 214 215 216 217 218 219
		var err error
		ds, err = components.NewDataCoord(ctx, factory)
		if err != nil {
			panic(err)
		}
		wg.Done()
		_ = ds.Run()
D
dragondriver 已提交
220 221 222
		if !localMsg {
			http.Handle(healthz.HealthzRouterPath, &componentsHealthzHandler{component: ds})
		}
223 224 225 226 227 228 229
	}()
	wg.Wait()

	metrics.RegisterDataCoord()
	return ds
}

X
XuanYang-cn 已提交
230
func (mr *MilvusRoles) runDataNode(ctx context.Context, localMsg bool, alias string) *components.DataNode {
231 232 233 234 235
	var dn *components.DataNode
	var wg sync.WaitGroup

	wg.Add(1)
	go func() {
X
XuanYang-cn 已提交
236
		datanode.Params.InitAlias(alias)
237
		datanode.Params.Init()
X
Xiaofan 已提交
238 239
		f := setLoggerFunc(localMsg)
		datanode.Params.SetLogConfig(f)
Y
yukun 已提交
240
		factory := newMsgFactory(localMsg)
241 242 243 244 245 246 247
		var err error
		dn, err = components.NewDataNode(ctx, factory)
		if err != nil {
			panic(err)
		}
		wg.Done()
		_ = dn.Run()
D
dragondriver 已提交
248 249 250
		if !localMsg {
			http.Handle(healthz.HealthzRouterPath, &componentsHealthzHandler{component: dn})
		}
251 252 253 254 255 256 257 258 259 260 261 262 263
	}()
	wg.Wait()

	metrics.RegisterDataNode()
	return dn
}

func (mr *MilvusRoles) runIndexCoord(ctx context.Context, localMsg bool) *components.IndexCoord {
	var is *components.IndexCoord
	var wg sync.WaitGroup

	wg.Add(1)
	go func() {
264
		indexcoord.Params.Init()
265

X
Xiaofan 已提交
266 267
		f := setLoggerFunc(localMsg)
		indexcoord.Params.SetLogConfig(f)
268 269 270 271 272 273 274
		var err error
		is, err = components.NewIndexCoord(ctx)
		if err != nil {
			panic(err)
		}
		wg.Done()
		_ = is.Run()
D
dragondriver 已提交
275 276 277
		if !localMsg {
			http.Handle(healthz.HealthzRouterPath, &componentsHealthzHandler{component: is})
		}
278 279 280 281 282 283 284
	}()
	wg.Wait()

	metrics.RegisterIndexCoord()
	return is
}

X
XuanYang-cn 已提交
285
func (mr *MilvusRoles) runIndexNode(ctx context.Context, localMsg bool, alias string) *components.IndexNode {
286 287 288 289 290
	var in *components.IndexNode
	var wg sync.WaitGroup

	wg.Add(1)
	go func() {
X
XuanYang-cn 已提交
291
		indexnode.Params.InitAlias(alias)
292 293
		indexnode.Params.Init()

X
Xiaofan 已提交
294 295
		f := setLoggerFunc(localMsg)
		indexnode.Params.SetLogConfig(f)
296 297 298 299 300 301 302
		var err error
		in, err = components.NewIndexNode(ctx)
		if err != nil {
			panic(err)
		}
		wg.Done()
		_ = in.Run()
D
dragondriver 已提交
303 304 305
		if !localMsg {
			http.Handle(healthz.HealthzRouterPath, &componentsHealthzHandler{component: in})
		}
306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332
	}()
	wg.Wait()

	metrics.RegisterIndexNode()
	return in
}

func (mr *MilvusRoles) runMsgStreamCoord(ctx context.Context) *components.MsgStream {
	var mss *components.MsgStream
	var wg sync.WaitGroup

	wg.Add(1)
	go func() {
		var err error
		mss, err = components.NewMsgStreamCoord(ctx)
		if err != nil {
			panic(err)
		}
		wg.Done()
		_ = mss.Run()
	}()
	wg.Wait()

	metrics.RegisterMsgStreamCoord()
	return mss
}

X
XuanYang-cn 已提交
333
func (mr *MilvusRoles) Run(localMsg bool, alias string) {
334
	if os.Getenv(metricsinfo.DeployModeEnvKey) == metricsinfo.StandaloneDeployMode {
335 336 337 338 339 340 341 342 343 344
		closer := trace.InitTracing("standalone")
		if closer != nil {
			defer closer.Close()
		}
	}

	ctx, cancel := context.WithCancel(context.Background())

	// only standalone enable localMsg
	if localMsg {
345
		os.Setenv(metricsinfo.DeployModeEnvKey, metricsinfo.StandaloneDeployMode)
346 347 348
		cfg := mr.setLogConfigFilename("standalone.log")
		logutil.SetupLogger(cfg)
		defer log.Sync()
349 350 351
	} else {
		err := os.Setenv(metricsinfo.DeployModeEnvKey, metricsinfo.ClusterDeployMode)
		if err != nil {
S
sunby 已提交
352
			log.Error("failed to set deploy mode: ", zap.Error(err))
353
		}
354 355 356 357 358 359 360 361 362 363 364 365
	}

	var rc *components.RootCoord
	if mr.EnableRootCoord {
		rc = mr.runRootCoord(ctx, localMsg)
		if rc != nil {
			defer rc.Stop()
		}
	}

	var pn *components.Proxy
	if mr.EnableProxy {
X
XuanYang-cn 已提交
366
		pn = mr.runProxy(ctx, localMsg, alias)
367 368 369 370 371 372 373 374 375 376 377 378 379 380 381
		if pn != nil {
			defer pn.Stop()
		}
	}

	var qs *components.QueryCoord
	if mr.EnableQueryCoord {
		qs = mr.runQueryCoord(ctx, localMsg)
		if qs != nil {
			defer qs.Stop()
		}
	}

	var qn *components.QueryNode
	if mr.EnableQueryNode {
X
XuanYang-cn 已提交
382
		qn = mr.runQueryNode(ctx, localMsg, alias)
383 384 385 386 387 388 389 390 391 392 393 394 395 396 397
		if qn != nil {
			defer qn.Stop()
		}
	}

	var ds *components.DataCoord
	if mr.EnableDataCoord {
		ds = mr.runDataCoord(ctx, localMsg)
		if ds != nil {
			defer ds.Stop()
		}
	}

	var dn *components.DataNode
	if mr.EnableDataNode {
X
XuanYang-cn 已提交
398
		dn = mr.runDataNode(ctx, localMsg, alias)
399 400 401 402 403 404 405 406 407 408 409 410 411 412 413
		if dn != nil {
			defer dn.Stop()
		}
	}

	var is *components.IndexCoord
	if mr.EnableIndexCoord {
		is = mr.runIndexCoord(ctx, localMsg)
		if is != nil {
			defer is.Stop()
		}
	}

	var in *components.IndexNode
	if mr.EnableIndexNode {
X
XuanYang-cn 已提交
414
		in = mr.runIndexNode(ctx, localMsg, alias)
415 416 417 418 419 420 421 422 423 424 425 426 427
		if in != nil {
			defer in.Stop()
		}
	}

	var mss *components.MsgStream
	if mr.EnableMsgStreamCoord {
		mss = mr.runMsgStreamCoord(ctx)
		if mss != nil {
			defer mss.Stop()
		}
	}

D
dragondriver 已提交
428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460
	if localMsg {
		standaloneHealthzHandler := func(w http.ResponseWriter, r *http.Request) {
			if rc == nil {
				rootCoordNotServingHandler(w, r)
				return
			}
			if qs == nil {
				queryCoordNotServingHandler(w, r)
				return
			}
			if ds == nil {
				dataCoordNotServingHandler(w, r)
				return
			}
			if is == nil {
				indexCoordNotServingHandler(w, r)
				return
			}
			// TODO(dragondriver): need to check node state?

			w.WriteHeader(http.StatusOK)
			w.Header().Set(healthz.ContentTypeHeader, healthz.ContentTypeText)
			_, err := fmt.Fprint(w, "OK")
			if err != nil {
				log.Warn("failed to send response",
					zap.Error(err))
			}

			// TODO(dragondriver): handle component states
		}
		http.HandleFunc(healthz.HealthzRouterPath, standaloneHealthzHandler)
	}

461 462 463 464 465 466 467 468 469
	metrics.ServeHTTP()

	sc := make(chan os.Signal, 1)
	signal.Notify(sc,
		syscall.SIGHUP,
		syscall.SIGINT,
		syscall.SIGTERM,
		syscall.SIGQUIT)
	sig := <-sc
S
sunby 已提交
470
	log.Error("Get signal to exit\n", zap.String("signal", sig.String()))
471 472 473 474

	// some deferred Stop has race with context cancel
	cancel()
}
X
Xiaofan 已提交
475 476 477 478 479 480 481 482 483 484 485 486

func setLoggerFunc(localMsg bool) func(cfg log.Config) {
	if !localMsg {
		return func(cfg log.Config) {
			log.Info("Set log file to ", zap.String("path", cfg.File.Filename))
			logutil.SetupLogger(&cfg)
			defer log.Sync()
		}
	}
	// no need to setup logger for standalone
	return func(cfg log.Config) {}
}