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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582
|
// Copyright 2016 The etcd Authors
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// 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.
package embed
import (
"context"
"crypto/tls"
"fmt"
"io/ioutil"
defaultLog "log"
"net"
"net/http"
"net/url"
"strconv"
"sync"
"time"
"github.com/coreos/etcd/compactor"
"github.com/coreos/etcd/etcdserver"
"github.com/coreos/etcd/etcdserver/api/etcdhttp"
"github.com/coreos/etcd/etcdserver/api/v2http"
"github.com/coreos/etcd/etcdserver/api/v2v3"
"github.com/coreos/etcd/etcdserver/api/v3client"
"github.com/coreos/etcd/etcdserver/api/v3rpc"
"github.com/coreos/etcd/pkg/cors"
"github.com/coreos/etcd/pkg/debugutil"
runtimeutil "github.com/coreos/etcd/pkg/runtime"
"github.com/coreos/etcd/pkg/transport"
"github.com/coreos/etcd/pkg/types"
"github.com/coreos/etcd/rafthttp"
"github.com/coreos/pkg/capnslog"
grpc_prometheus "github.com/grpc-ecosystem/go-grpc-prometheus"
"github.com/soheilhy/cmux"
"google.golang.org/grpc"
"google.golang.org/grpc/keepalive"
)
var plog = capnslog.NewPackageLogger("github.com/coreos/etcd", "embed")
const (
// internal fd usage includes disk usage and transport usage.
// To read/write snapshot, snap pkg needs 1. In normal case, wal pkg needs
// at most 2 to read/lock/write WALs. One case that it needs to 2 is to
// read all logs after some snapshot index, which locates at the end of
// the second last and the head of the last. For purging, it needs to read
// directory, so it needs 1. For fd monitor, it needs 1.
// For transport, rafthttp builds two long-polling connections and at most
// four temporary connections with each member. There are at most 9 members
// in a cluster, so it should reserve 96.
// For the safety, we set the total reserved number to 150.
reservedInternalFDNum = 150
)
// Etcd contains a running etcd server and its listeners.
type Etcd struct {
Peers []*peerListener
Clients []net.Listener
// a map of contexts for the servers that serves client requests.
sctxs map[string]*serveCtx
metricsListeners []net.Listener
Server *etcdserver.EtcdServer
cfg Config
stopc chan struct{}
errc chan error
closeOnce sync.Once
}
type peerListener struct {
net.Listener
serve func() error
close func(context.Context) error
}
// StartEtcd launches the etcd server and HTTP handlers for client/server communication.
// The returned Etcd.Server is not guaranteed to have joined the cluster. Wait
// on the Etcd.Server.ReadyNotify() channel to know when it completes and is ready for use.
func StartEtcd(inCfg *Config) (e *Etcd, err error) {
if err = inCfg.Validate(); err != nil {
return nil, err
}
serving := false
e = &Etcd{cfg: *inCfg, stopc: make(chan struct{})}
cfg := &e.cfg
defer func() {
if e == nil || err == nil {
return
}
if !serving {
// errored before starting gRPC server for serveCtx.serversC
for _, sctx := range e.sctxs {
close(sctx.serversC)
}
}
e.Close()
e = nil
}()
if e.Peers, err = startPeerListeners(cfg); err != nil {
return e, err
}
if e.sctxs, err = startClientListeners(cfg); err != nil {
return e, err
}
for _, sctx := range e.sctxs {
e.Clients = append(e.Clients, sctx.l)
}
var (
urlsmap types.URLsMap
token string
)
memberInitialized := true
if !isMemberInitialized(cfg) {
memberInitialized = false
urlsmap, token, err = cfg.PeerURLsMapAndToken("etcd")
if err != nil {
return e, fmt.Errorf("error setting up initial cluster: %v", err)
}
}
// AutoCompactionRetention defaults to "0" if not set.
if len(cfg.AutoCompactionRetention) == 0 {
cfg.AutoCompactionRetention = "0"
}
autoCompactionRetention, err := parseCompactionRetention(cfg.AutoCompactionMode, cfg.AutoCompactionRetention)
if err != nil {
return e, err
}
srvcfg := etcdserver.ServerConfig{
Name: cfg.Name,
ClientURLs: cfg.ACUrls,
PeerURLs: cfg.APUrls,
DataDir: cfg.Dir,
DedicatedWALDir: cfg.WalDir,
SnapCount: cfg.SnapCount,
MaxSnapFiles: cfg.MaxSnapFiles,
MaxWALFiles: cfg.MaxWalFiles,
InitialPeerURLsMap: urlsmap,
InitialClusterToken: token,
DiscoveryURL: cfg.Durl,
DiscoveryProxy: cfg.Dproxy,
NewCluster: cfg.IsNewCluster(),
ForceNewCluster: cfg.ForceNewCluster,
PeerTLSInfo: cfg.PeerTLSInfo,
TickMs: cfg.TickMs,
ElectionTicks: cfg.ElectionTicks(),
InitialElectionTickAdvance: cfg.InitialElectionTickAdvance,
AutoCompactionRetention: autoCompactionRetention,
AutoCompactionMode: cfg.AutoCompactionMode,
QuotaBackendBytes: cfg.QuotaBackendBytes,
MaxTxnOps: cfg.MaxTxnOps,
MaxRequestBytes: cfg.MaxRequestBytes,
StrictReconfigCheck: cfg.StrictReconfigCheck,
ClientCertAuthEnabled: cfg.ClientTLSInfo.ClientCertAuth,
AuthToken: cfg.AuthToken,
InitialCorruptCheck: cfg.ExperimentalInitialCorruptCheck,
CorruptCheckTime: cfg.ExperimentalCorruptCheckTime,
Debug: cfg.Debug,
}
if e.Server, err = etcdserver.NewServer(srvcfg); err != nil {
return e, err
}
// buffer channel so goroutines on closed connections won't wait forever
e.errc = make(chan error, len(e.Peers)+len(e.Clients)+2*len(e.sctxs))
// newly started member ("memberInitialized==false")
// does not need corruption check
if memberInitialized {
if err = e.Server.CheckInitialHashKV(); err != nil {
// set "EtcdServer" to nil, so that it does not block on "EtcdServer.Close()"
// (nothing to close since rafthttp transports have not been started)
e.Server = nil
return e, err
}
}
e.Server.Start()
if err = e.servePeers(); err != nil {
return e, err
}
if err = e.serveClients(); err != nil {
return e, err
}
if err = e.serveMetrics(); err != nil {
return e, err
}
serving = true
return e, nil
}
// Config returns the current configuration.
func (e *Etcd) Config() Config {
return e.cfg
}
// Close gracefully shuts down all servers/listeners.
// Client requests will be terminated with request timeout.
// After timeout, enforce remaning requests be closed immediately.
func (e *Etcd) Close() {
e.closeOnce.Do(func() { close(e.stopc) })
// close client requests with request timeout
timeout := 2 * time.Second
if e.Server != nil {
timeout = e.Server.Cfg.ReqTimeout()
}
for _, sctx := range e.sctxs {
for ss := range sctx.serversC {
ctx, cancel := context.WithTimeout(context.Background(), timeout)
stopServers(ctx, ss)
cancel()
}
}
for _, sctx := range e.sctxs {
sctx.cancel()
}
for i := range e.Clients {
if e.Clients[i] != nil {
e.Clients[i].Close()
}
}
for i := range e.metricsListeners {
e.metricsListeners[i].Close()
}
// close rafthttp transports
if e.Server != nil {
e.Server.Stop()
}
// close all idle connections in peer handler (wait up to 1-second)
for i := range e.Peers {
if e.Peers[i] != nil && e.Peers[i].close != nil {
ctx, cancel := context.WithTimeout(context.Background(), time.Second)
e.Peers[i].close(ctx)
cancel()
}
}
}
func stopServers(ctx context.Context, ss *servers) {
shutdownNow := func() {
// first, close the http.Server
ss.http.Shutdown(ctx)
// then close grpc.Server; cancels all active RPCs
ss.grpc.Stop()
}
// do not grpc.Server.GracefulStop with TLS enabled etcd server
// See https://github.com/grpc/grpc-go/issues/1384#issuecomment-317124531
// and https://github.com/coreos/etcd/issues/8916
if ss.secure {
shutdownNow()
return
}
ch := make(chan struct{})
go func() {
defer close(ch)
// close listeners to stop accepting new connections,
// will block on any existing transports
ss.grpc.GracefulStop()
}()
// wait until all pending RPCs are finished
select {
case <-ch:
case <-ctx.Done():
// took too long, manually close open transports
// e.g. watch streams
shutdownNow()
// concurrent GracefulStop should be interrupted
<-ch
}
}
func (e *Etcd) Err() <-chan error { return e.errc }
func startPeerListeners(cfg *Config) (peers []*peerListener, err error) {
if err = updateCipherSuites(&cfg.PeerTLSInfo, cfg.CipherSuites); err != nil {
return nil, err
}
if err = cfg.PeerSelfCert(); err != nil {
plog.Fatalf("could not get certs (%v)", err)
}
if !cfg.PeerTLSInfo.Empty() {
plog.Infof("peerTLS: %s", cfg.PeerTLSInfo)
}
peers = make([]*peerListener, len(cfg.LPUrls))
defer func() {
if err == nil {
return
}
for i := range peers {
if peers[i] != nil && peers[i].close != nil {
plog.Info("stopping listening for peers on ", cfg.LPUrls[i].String())
ctx, cancel := context.WithTimeout(context.Background(), time.Second)
peers[i].close(ctx)
cancel()
}
}
}()
for i, u := range cfg.LPUrls {
if u.Scheme == "http" {
if !cfg.PeerTLSInfo.Empty() {
plog.Warningf("The scheme of peer url %s is HTTP while peer key/cert files are presented. Ignored peer key/cert files.", u.String())
}
if cfg.PeerTLSInfo.ClientCertAuth {
plog.Warningf("The scheme of peer url %s is HTTP while client cert auth (--peer-client-cert-auth) is enabled. Ignored client cert auth for this url.", u.String())
}
}
peers[i] = &peerListener{close: func(context.Context) error { return nil }}
peers[i].Listener, err = rafthttp.NewListener(u, &cfg.PeerTLSInfo)
if err != nil {
return nil, err
}
// once serve, overwrite with 'http.Server.Shutdown'
peers[i].close = func(context.Context) error {
return peers[i].Listener.Close()
}
plog.Info("listening for peers on ", u.String())
}
return peers, nil
}
// configure peer handlers after rafthttp.Transport started
func (e *Etcd) servePeers() (err error) {
ph := etcdhttp.NewPeerHandler(e.Server)
var peerTLScfg *tls.Config
if !e.cfg.PeerTLSInfo.Empty() {
if peerTLScfg, err = e.cfg.PeerTLSInfo.ServerConfig(); err != nil {
return err
}
}
for _, p := range e.Peers {
gs := v3rpc.Server(e.Server, peerTLScfg)
m := cmux.New(p.Listener)
go gs.Serve(m.Match(cmux.HTTP2()))
srv := &http.Server{
Handler: grpcHandlerFunc(gs, ph),
ReadTimeout: 5 * time.Minute,
ErrorLog: defaultLog.New(ioutil.Discard, "", 0), // do not log user error
}
go srv.Serve(m.Match(cmux.Any()))
p.serve = func() error { return m.Serve() }
p.close = func(ctx context.Context) error {
// gracefully shutdown http.Server
// close open listeners, idle connections
// until context cancel or time-out
stopServers(ctx, &servers{secure: peerTLScfg != nil, grpc: gs, http: srv})
return nil
}
}
// start peer servers in a goroutine
for _, pl := range e.Peers {
go func(l *peerListener) {
e.errHandler(l.serve())
}(pl)
}
return nil
}
func startClientListeners(cfg *Config) (sctxs map[string]*serveCtx, err error) {
if err = updateCipherSuites(&cfg.ClientTLSInfo, cfg.CipherSuites); err != nil {
return nil, err
}
if err = cfg.ClientSelfCert(); err != nil {
plog.Fatalf("could not get certs (%v)", err)
}
if cfg.EnablePprof {
plog.Infof("pprof is enabled under %s", debugutil.HTTPPrefixPProf)
}
sctxs = make(map[string]*serveCtx)
for _, u := range cfg.LCUrls {
sctx := newServeCtx()
if u.Scheme == "http" || u.Scheme == "unix" {
if !cfg.ClientTLSInfo.Empty() {
plog.Warningf("The scheme of client url %s is HTTP while peer key/cert files are presented. Ignored key/cert files.", u.String())
}
if cfg.ClientTLSInfo.ClientCertAuth {
plog.Warningf("The scheme of client url %s is HTTP while client cert auth (--client-cert-auth) is enabled. Ignored client cert auth for this url.", u.String())
}
}
if (u.Scheme == "https" || u.Scheme == "unixs") && cfg.ClientTLSInfo.Empty() {
return nil, fmt.Errorf("TLS key/cert (--cert-file, --key-file) must be provided for client url %s with HTTPs scheme", u.String())
}
proto := "tcp"
addr := u.Host
if u.Scheme == "unix" || u.Scheme == "unixs" {
proto = "unix"
addr = u.Host + u.Path
}
sctx.secure = u.Scheme == "https" || u.Scheme == "unixs"
sctx.insecure = !sctx.secure
if oldctx := sctxs[addr]; oldctx != nil {
oldctx.secure = oldctx.secure || sctx.secure
oldctx.insecure = oldctx.insecure || sctx.insecure
continue
}
if sctx.l, err = net.Listen(proto, addr); err != nil {
return nil, err
}
// net.Listener will rewrite ipv4 0.0.0.0 to ipv6 [::], breaking
// hosts that disable ipv6. So, use the address given by the user.
sctx.addr = addr
if fdLimit, fderr := runtimeutil.FDLimit(); fderr == nil {
if fdLimit <= reservedInternalFDNum {
plog.Fatalf("file descriptor limit[%d] of etcd process is too low, and should be set higher than %d to ensure internal usage", fdLimit, reservedInternalFDNum)
}
sctx.l = transport.LimitListener(sctx.l, int(fdLimit-reservedInternalFDNum))
}
if proto == "tcp" {
if sctx.l, err = transport.NewKeepAliveListener(sctx.l, "tcp", nil); err != nil {
return nil, err
}
}
plog.Info("listening for client requests on ", u.Host)
defer func() {
if err != nil {
sctx.l.Close()
plog.Info("stopping listening for client requests on ", u.Host)
}
}()
for k := range cfg.UserHandlers {
sctx.userHandlers[k] = cfg.UserHandlers[k]
}
sctx.serviceRegister = cfg.ServiceRegister
if cfg.EnablePprof || cfg.Debug {
sctx.registerPprof()
}
if cfg.Debug {
sctx.registerTrace()
}
sctxs[addr] = sctx
}
return sctxs, nil
}
func (e *Etcd) serveClients() (err error) {
if !e.cfg.ClientTLSInfo.Empty() {
plog.Infof("ClientTLS: %s", e.cfg.ClientTLSInfo)
}
if e.cfg.CorsInfo.String() != "" {
plog.Infof("cors = %s", e.cfg.CorsInfo)
}
// Start a client server goroutine for each listen address
var h http.Handler
if e.Config().EnableV2 {
if len(e.Config().ExperimentalEnableV2V3) > 0 {
srv := v2v3.NewServer(v3client.New(e.Server), e.cfg.ExperimentalEnableV2V3)
h = v2http.NewClientHandler(srv, e.Server.Cfg.ReqTimeout())
} else {
h = v2http.NewClientHandler(e.Server, e.Server.Cfg.ReqTimeout())
}
} else {
mux := http.NewServeMux()
etcdhttp.HandleBasic(mux, e.Server)
h = mux
}
h = http.Handler(&cors.CORSHandler{Handler: h, Info: e.cfg.CorsInfo})
gopts := []grpc.ServerOption{}
if e.cfg.GRPCKeepAliveMinTime > time.Duration(0) {
gopts = append(gopts, grpc.KeepaliveEnforcementPolicy(keepalive.EnforcementPolicy{
MinTime: e.cfg.GRPCKeepAliveMinTime,
PermitWithoutStream: false,
}))
}
if e.cfg.GRPCKeepAliveInterval > time.Duration(0) &&
e.cfg.GRPCKeepAliveTimeout > time.Duration(0) {
gopts = append(gopts, grpc.KeepaliveParams(keepalive.ServerParameters{
Time: e.cfg.GRPCKeepAliveInterval,
Timeout: e.cfg.GRPCKeepAliveTimeout,
}))
}
// start client servers in a goroutine
for _, sctx := range e.sctxs {
go func(s *serveCtx) {
e.errHandler(s.serve(e.Server, &e.cfg.ClientTLSInfo, h, e.errHandler, gopts...))
}(sctx)
}
return nil
}
func (e *Etcd) serveMetrics() (err error) {
if e.cfg.Metrics == "extensive" {
grpc_prometheus.EnableHandlingTimeHistogram()
}
if len(e.cfg.ListenMetricsUrls) > 0 {
metricsMux := http.NewServeMux()
etcdhttp.HandleMetricsHealth(metricsMux, e.Server)
for _, murl := range e.cfg.ListenMetricsUrls {
tlsInfo := &e.cfg.ClientTLSInfo
if murl.Scheme == "http" {
tlsInfo = nil
}
ml, err := transport.NewListener(murl.Host, murl.Scheme, tlsInfo)
if err != nil {
return err
}
e.metricsListeners = append(e.metricsListeners, ml)
go func(u url.URL, ln net.Listener) {
plog.Info("listening for metrics on ", u.String())
e.errHandler(http.Serve(ln, metricsMux))
}(murl, ml)
}
}
return nil
}
func (e *Etcd) errHandler(err error) {
select {
case <-e.stopc:
return
default:
}
select {
case <-e.stopc:
case e.errc <- err:
}
}
func parseCompactionRetention(mode, retention string) (ret time.Duration, err error) {
h, err := strconv.Atoi(retention)
if err == nil {
switch mode {
case compactor.ModeRevision:
ret = time.Duration(int64(h))
case compactor.ModePeriodic:
ret = time.Duration(int64(h)) * time.Hour
}
} else {
// periodic compaction
ret, err = time.ParseDuration(retention)
if err != nil {
return 0, fmt.Errorf("error parsing CompactionRetention: %v", err)
}
}
return ret, nil
}
|