TiDB 优雅关闭

#( 07/31/23@ 8:32下午 )( happy@ZBMAC-f298743e3 ):~/docker/tiup/tiproxy   tiup playground v6.4.0 --db 2 --kv 1 --pd 1 --tiflash 0 --without-monitor --db.config tidb.tomltiup is checking updates for component playground ...Starting component `playground`: /Users/happy/.tiup/components/playground/v1.12.5/tiup-playground v6.4.0 --db 2 --kv 1 --pd 1 --tiflash 0 --without-monitor --db.config tidb.tomlStart pd instance:v6.4.0Start tikv instance:v6.4.0Start tidb instance:v6.4.0Start tidb instance:v6.4.0Waiting for tidb instances ready127.0.0.1:4000 ... Done127.0.0.1:4001 ... Done
  TiDB Playground Cluster is started, enjoy!
Connect TiDB:   mysql --comments --host 127.0.0.1 --port 4000 -u rootConnect TiDB:   mysql --comments --host 127.0.0.1 --port 4001 -u rootTiDB Dashboard: http://127.0.0.1:2379/dashboard

#( 07/31/23@ 8:32下午 )( happy@ZBMAC-f298743e3 ):~   tiup playground displaytiup is checking updates for component playground ...Starting component `playground`: /Users/happy/.tiup/components/playground/v1.12.5/tiup-playground displayPid    Role  Uptime---    ----  ------10113  pd    49.376485092s10114  tikv  49.32262974s10115  tidb  49.283144092s10116  tidb  49.245069308s

#( 07/31/23@ 8:34下午 )( happy@ZBMAC-f298743e3 ):~   tiup playground scale-in --pid 10115tiup is checking updates for component playground ...Starting component `playground`: /Users/happy/.tiup/components/playground/v1.12.5/tiup-playground scale-in --pid 10115scale in tidb success

#( 07/31/23@ 8:34下午 )( happy@ZBMAC-f298743e3 ):~   ps -ef | grep 10115  502 11371 99718   0  8:34下午 ttys001    0:00.00 grep --color=auto --exclude-dir=.bzr --exclude-dir=CVS --exclude-dir=.git --exclude-dir=.hg --exclude-dir=.svn --exclude-dir=.idea --exclude-dir=.tox 10115  502 10115 10111   0  8:32下午 ttys005    0:04.29 /Users/happy/.tiup/components/tidb/v6.4.0/tidb-server -P 4000 --store=tikv --host=127.0.0.1 --status=10080 --path=127.0.0.1:2379 --log-file=/Users/happy/.tiup/data/TlaeoSj/tidb-0/tidb.log --config=/Users/happy/.tiup/data/TlaeoSj/tidb-0/tidb.toml

func main() {    //..    signal.SetupSignalHandler(func(graceful bool) {        svr.Close()        cleanup(svr, storage, dom, graceful)        cpuprofile.StopCPUProfiler()        close(exited)    })    // ...}

// Close closes the server.func (s *Server) Close() {    s.startShutdown()    s.rwlock.Lock() // prevent new connections    defer s.rwlock.Unlock()
    if s.listener != nil {        err := s.listener.Close()        terror.Log(errors.Trace(err))        s.listener = nil    }    if s.socket != nil {        err := s.socket.Close()        terror.Log(errors.Trace(err))        s.socket = nil    }    if s.statusServer != nil {        err := s.statusServer.Close()        terror.Log(errors.Trace(err))        s.statusServer = nil    }    if s.grpcServer != nil {        s.grpcServer.Stop()        s.grpcServer = nil    }    if s.autoIDService != nil {        s.autoIDService.Close()    }    if s.authTokenCancelFunc != nil {        s.authTokenCancelFunc()    }    s.wg.Wait()    metrics.ServerEventCounter.WithLabelValues(metrics.EventClose).Inc()}
func (s *Server) startShutdown() {    s.rwlock.RLock()    logutil.BgLogger().Info("setting tidb-server to report unhealthy (shutting-down)")    s.inShutdownMode = true    s.rwlock.RUnlock()    // give the load balancer a chance to receive a few unhealthy health reports    // before acquiring the s.rwlock and blocking connections.    waitTime := time.Duration(s.cfg.GracefulWaitBeforeShutdown) * time.Second    if waitTime > 0 {        logutil.BgLogger().Info("waiting for stray connections before starting shutdown process", zap.Duration("waitTime", waitTime))        time.Sleep(waitTime)    }}

func cleanup(svr *server.Server, storage kv.Storage, dom *domain.Domain, graceful bool) {    if graceful {        done := make(chan struct{})        svr.GracefulDown(context.Background(), done)    } else {        svr.TryGracefulDown()    }    plugin.Shutdown(context.Background())    closeDomainAndStorage(storage, dom)    disk.CleanUp()    topsql.Close()}

var gracefulCloseConnectionsTimeout = 15 * time.Second
// TryGracefulDown will try to gracefully close all connection first with timeout. if timeout, will close all connection directly.func (s *Server) TryGracefulDown() {    ctx, cancel := context.WithTimeout(context.Background(), gracefulCloseConnectionsTimeout)    defer cancel()    done := make(chan struct{})    go func() {        s.GracefulDown(ctx, done)    }()    select {    case <-ctx.Done():        s.KillAllConnections()    case <-done:        return    }}

// GracefulDown waits all clients to close.func (s *Server) GracefulDown(ctx context.Context, done chan struct{}) {    logutil.Logger(ctx).Info("[server] graceful shutdown.")    metrics.ServerEventCounter.WithLabelValues(metrics.EventGracefulDown).Inc()
    count := s.ConnectionCount()    for i := 0; count > 0; i++ {        s.kickIdleConnection()
        count = s.ConnectionCount()        if count == 0 {            break        }        // Print information for every 30s.        if i%30 == 0 {            logutil.Logger(ctx).Info("graceful shutdown...", zap.Int("conn count", count))        }        ticker := time.After(time.Second)        select {        case <-ctx.Done():            return        case <-ticker:        }    }    close(done)}

// ConnectionCount gets current connection count.func (s *Server) ConnectionCount() int {    s.rwlock.RLock()    cnt := len(s.clients)    s.rwlock.RUnlock()    return cnt}

func (s *Server) kickIdleConnection() {    var conns []*clientConn    s.rwlock.RLock()    for _, cc := range s.clients {        if cc.ShutdownOrNotify() {            // Shutdowned conn will be closed by us, and notified conn will exist themselves.            conns = append(conns, cc)        }    }    s.rwlock.RUnlock()
    for _, cc := range conns {        err := cc.Close()        if err != nil {            logutil.BgLogger().Error("close connection", zap.Error(err))        }    }}

mysql> show processlist;+---------------------+------+-----------------+------+---------+------+----------------------------+------------------+| Id                  | User | Host            | db   | Command | Time | State                      | Info             |+---------------------+------+-----------------+------+---------+------+----------------------------+------------------+| 7794237818187809175 | root | 127.0.0.1:61293 | a    | Query   |    0 | in transaction; autocommit | show processlist |+---------------------+------+-----------------+------+---------+------+----------------------------+------------------+1 row in set (0.00 sec)