mirror of
https://github.com/HDT3213/godis.git
synced 2025-10-05 00:42:43 +08:00
403 lines
11 KiB
Go
403 lines
11 KiB
Go
package database
|
|
|
|
import (
|
|
"bytes"
|
|
"context"
|
|
"errors"
|
|
"fmt"
|
|
"github.com/hdt3213/godis/config"
|
|
"github.com/hdt3213/godis/interface/redis"
|
|
"github.com/hdt3213/godis/lib/logger"
|
|
"github.com/hdt3213/godis/lib/utils"
|
|
"github.com/hdt3213/godis/redis/connection"
|
|
"github.com/hdt3213/godis/redis/parser"
|
|
"github.com/hdt3213/godis/redis/protocol"
|
|
rdb "github.com/hdt3213/rdb/parser"
|
|
"net"
|
|
"strconv"
|
|
"strings"
|
|
"sync"
|
|
"sync/atomic"
|
|
"time"
|
|
)
|
|
|
|
const (
|
|
masterRole = iota
|
|
slaveRole
|
|
)
|
|
|
|
type replicationStatus struct {
|
|
mutex sync.Mutex
|
|
ctx context.Context
|
|
cancel context.CancelFunc
|
|
modCount int32 // execSlaveOf() or reconnect will cause modCount++
|
|
|
|
masterHost string
|
|
masterPort int
|
|
|
|
masterConn net.Conn
|
|
masterChan <-chan *parser.Payload
|
|
replId string
|
|
replOffset int64
|
|
lastRecvTime time.Time
|
|
running sync.WaitGroup
|
|
}
|
|
|
|
func initReplStatus() *replicationStatus {
|
|
repl := &replicationStatus{}
|
|
// start cron
|
|
return repl
|
|
}
|
|
|
|
func (mdb *MultiDB) startReplCron() {
|
|
go func() {
|
|
defer func() {
|
|
if err := recover(); err != nil {
|
|
logger.Error("panic", err)
|
|
}
|
|
}()
|
|
ticker := time.Tick(time.Second)
|
|
for range ticker {
|
|
mdb.slaveCron()
|
|
}
|
|
}()
|
|
}
|
|
|
|
func (mdb *MultiDB) execSlaveOf(c redis.Connection, args [][]byte) redis.Reply {
|
|
if strings.ToLower(string(args[0])) == "no" &&
|
|
strings.ToLower(string(args[1])) == "one" {
|
|
mdb.slaveOfNone()
|
|
return protocol.MakeOkReply()
|
|
}
|
|
host := string(args[0])
|
|
port, err := strconv.Atoi(string(args[1]))
|
|
if err != nil {
|
|
return protocol.MakeErrReply("ERR value is not an integer or out of range")
|
|
}
|
|
mdb.replication.mutex.Lock()
|
|
atomic.StoreInt32(&mdb.role, slaveRole)
|
|
mdb.replication.masterHost = host
|
|
mdb.replication.masterPort = port
|
|
// use buffered channel in case receiver goroutine exited before controller send stop signal
|
|
atomic.AddInt32(&mdb.replication.modCount, 1)
|
|
mdb.replication.mutex.Unlock()
|
|
go mdb.syncWithMaster()
|
|
return protocol.MakeOkReply()
|
|
}
|
|
|
|
func (mdb *MultiDB) slaveOfNone() {
|
|
mdb.replication.mutex.Lock()
|
|
defer mdb.replication.mutex.Unlock()
|
|
mdb.replication.masterHost = ""
|
|
mdb.replication.masterPort = 0
|
|
mdb.replication.replId = ""
|
|
mdb.replication.replOffset = -1
|
|
mdb.replication.stopSlaveWithMutex()
|
|
}
|
|
|
|
// stopSlaveWithMutex stops in-progress connectWithMaster/fullSync/receiveAOF
|
|
// invoker should have replication mutex
|
|
func (repl *replicationStatus) stopSlaveWithMutex() {
|
|
// update modCount to stop connectWithMaster and fullSync
|
|
atomic.AddInt32(&repl.modCount, 1)
|
|
// send cancel to receiveAOF
|
|
if repl.cancel != nil {
|
|
repl.cancel()
|
|
repl.running.Wait()
|
|
}
|
|
repl.ctx = context.Background()
|
|
repl.cancel = nil
|
|
if repl.masterConn != nil {
|
|
_ = repl.masterConn.Close() // parser.ParseStream will close masterChan
|
|
}
|
|
repl.masterConn = nil
|
|
repl.masterChan = nil
|
|
}
|
|
|
|
func (repl *replicationStatus) close() error {
|
|
repl.mutex.Lock()
|
|
defer repl.mutex.Unlock()
|
|
repl.stopSlaveWithMutex()
|
|
return nil
|
|
}
|
|
|
|
func (mdb *MultiDB) syncWithMaster() {
|
|
defer func() {
|
|
if err := recover(); err != nil {
|
|
logger.Error(err)
|
|
}
|
|
}()
|
|
mdb.replication.mutex.Lock()
|
|
ctx, cancel := context.WithCancel(context.Background())
|
|
mdb.replication.ctx = ctx
|
|
mdb.replication.cancel = cancel
|
|
mdb.replication.mutex.Unlock()
|
|
err := mdb.connectWithMaster()
|
|
if err != nil {
|
|
// full sync failed, abort
|
|
return
|
|
}
|
|
err = mdb.doPsync()
|
|
if err != nil {
|
|
// full sync failed, abort
|
|
return
|
|
}
|
|
err = mdb.receiveAOF()
|
|
if err != nil {
|
|
// full sync failed, abort
|
|
return
|
|
}
|
|
}
|
|
|
|
func (mdb *MultiDB) connectWithMaster() error {
|
|
modCount := atomic.LoadInt32(&mdb.replication.modCount)
|
|
addr := mdb.replication.masterHost + ":" + strconv.Itoa(mdb.replication.masterPort)
|
|
conn, err := net.Dial("tcp", addr)
|
|
if err != nil {
|
|
mdb.slaveOfNone() // abort
|
|
return errors.New("connect master failed " + err.Error())
|
|
}
|
|
masterChan := parser.ParseStream(conn)
|
|
|
|
// ping
|
|
pingCmdLine := utils.ToCmdLine("ping")
|
|
pingReq := protocol.MakeMultiBulkReply(pingCmdLine)
|
|
_, err = conn.Write(pingReq.ToBytes())
|
|
if err != nil {
|
|
return errors.New("send failed " + err.Error())
|
|
}
|
|
pingResp := <-masterChan
|
|
if pingResp.Err != nil {
|
|
return errors.New("read response failed: " + pingResp.Err.Error())
|
|
}
|
|
switch reply := pingResp.Data.(type) {
|
|
case *protocol.StandardErrReply:
|
|
if !strings.HasPrefix(reply.Error(), "NOAUTH") &&
|
|
!strings.HasPrefix(reply.Error(), "NOPERM") &&
|
|
!strings.HasPrefix(reply.Error(), "ERR operation not permitted") {
|
|
logger.Error("Error reply to PING from master: " + string(reply.ToBytes()))
|
|
mdb.slaveOfNone() // abort
|
|
return nil
|
|
}
|
|
}
|
|
|
|
// just to reduce duplication of code
|
|
sendCmdToMaster := func(conn net.Conn, cmdLine CmdLine, masterChan <-chan *parser.Payload) error {
|
|
req := protocol.MakeMultiBulkReply(cmdLine)
|
|
_, err := conn.Write(req.ToBytes())
|
|
if err != nil {
|
|
mdb.slaveOfNone() // abort
|
|
return errors.New("send failed " + err.Error())
|
|
}
|
|
resp := <-masterChan
|
|
if resp.Err != nil {
|
|
mdb.slaveOfNone() // abort
|
|
return errors.New("read response failed: " + resp.Err.Error())
|
|
}
|
|
if !protocol.IsOKReply(resp.Data) {
|
|
mdb.slaveOfNone() // abort
|
|
return errors.New("unexpected auth response: " + string(resp.Data.ToBytes()))
|
|
}
|
|
return nil
|
|
}
|
|
|
|
// auth
|
|
if config.Properties.MasterAuth != "" {
|
|
authCmdLine := utils.ToCmdLine("auth", config.Properties.MasterAuth)
|
|
err = sendCmdToMaster(conn, authCmdLine, masterChan)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
}
|
|
|
|
// announce port
|
|
var port int
|
|
if config.Properties.SlaveAnnouncePort != 0 {
|
|
port = config.Properties.SlaveAnnouncePort
|
|
} else {
|
|
port = config.Properties.Port
|
|
}
|
|
portCmdLine := utils.ToCmdLine("REPLCONF", "listening-port", strconv.Itoa(port))
|
|
err = sendCmdToMaster(conn, portCmdLine, masterChan)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
|
|
// announce ip
|
|
if config.Properties.SlaveAnnounceIP != "" {
|
|
ipCmdLine := utils.ToCmdLine("REPLCONF", "ip-address", config.Properties.SlaveAnnounceIP)
|
|
err = sendCmdToMaster(conn, ipCmdLine, masterChan)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
}
|
|
|
|
// announce capacity
|
|
capaCmdLine := utils.ToCmdLine("REPLCONF", "capa", "psync2")
|
|
err = sendCmdToMaster(conn, capaCmdLine, masterChan)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
|
|
// update connection
|
|
mdb.replication.mutex.Lock()
|
|
if mdb.replication.modCount != modCount {
|
|
// replication conf changed during connecting and waiting mutex
|
|
return nil
|
|
}
|
|
mdb.replication.masterConn = conn
|
|
mdb.replication.masterChan = masterChan
|
|
mdb.replication.mutex.Unlock()
|
|
return nil
|
|
}
|
|
|
|
// doPsync send psync to master and sync repl-id then load rdb
|
|
func (mdb *MultiDB) doPsync() error {
|
|
modCount := atomic.LoadInt32(&mdb.replication.modCount)
|
|
psyncCmdLine := utils.ToCmdLine("psync", "?", "-1")
|
|
psyncReq := protocol.MakeMultiBulkReply(psyncCmdLine)
|
|
_, err := mdb.replication.masterConn.Write(psyncReq.ToBytes())
|
|
if err != nil {
|
|
return errors.New("send failed " + err.Error())
|
|
}
|
|
ch := mdb.replication.masterChan
|
|
psyncPayload1 := <-ch
|
|
if psyncPayload1.Err != nil {
|
|
return errors.New("read response failed: " + psyncPayload1.Err.Error())
|
|
}
|
|
psyncHeader, ok := psyncPayload1.Data.(*protocol.StatusReply)
|
|
if !ok {
|
|
return errors.New("illegal payload header: " + string(psyncPayload1.Data.ToBytes()))
|
|
}
|
|
headers := strings.Split(psyncHeader.Status, " ")
|
|
if len(headers) != 3 {
|
|
return errors.New("illegal payload header: " + psyncHeader.Status)
|
|
}
|
|
logger.Info("receive psync header from master")
|
|
psyncPayload2 := <-ch
|
|
if psyncPayload2.Err != nil {
|
|
return errors.New("read response failed: " + psyncPayload2.Err.Error())
|
|
}
|
|
psyncData, ok := psyncPayload2.Data.(*protocol.BulkReply)
|
|
if !ok {
|
|
return errors.New("illegal payload header: " + string(psyncPayload2.Data.ToBytes()))
|
|
}
|
|
logger.Info(fmt.Sprintf("receive %d bytes of rdb from master", len(psyncData.Arg)))
|
|
rdbDec := rdb.NewDecoder(bytes.NewReader(psyncData.Arg))
|
|
rdbHolder := MakeBasicMultiDB()
|
|
err = dumpRDB(rdbDec, rdbHolder)
|
|
if err != nil {
|
|
return errors.New("dump rdb failed: " + err.Error())
|
|
}
|
|
|
|
mdb.replication.mutex.Lock()
|
|
defer mdb.replication.mutex.Unlock()
|
|
if mdb.replication.modCount != modCount {
|
|
// replication conf changed during connecting and waiting mutex
|
|
return nil
|
|
}
|
|
mdb.replication.replId = headers[1]
|
|
mdb.replication.replOffset, err = strconv.ParseInt(headers[2], 10, 64)
|
|
if err != nil {
|
|
return errors.New("get illegal repl offset: " + headers[2])
|
|
}
|
|
logger.Info("full resync from master: " + mdb.replication.replId)
|
|
logger.Info("current offset:", mdb.replication.replOffset)
|
|
for i, newDB := range rdbHolder.dbSet {
|
|
oldDB := mdb.selectDB(i)
|
|
oldDB.Load(newDB)
|
|
}
|
|
// there is no CRLF between RDB and following AOF, reset stream to avoid parser error
|
|
mdb.replication.masterChan = parser.ParseStream(mdb.replication.masterConn)
|
|
// fixme: update aof file
|
|
return nil
|
|
}
|
|
|
|
func (mdb *MultiDB) receiveAOF() error {
|
|
conn := connection.NewConn(mdb.replication.masterConn)
|
|
conn.SetRole(connection.ReplicationRecvCli)
|
|
mdb.replication.mutex.Lock()
|
|
modCount := mdb.replication.modCount
|
|
done := mdb.replication.ctx.Done()
|
|
mdb.replication.mutex.Unlock()
|
|
if done == nil {
|
|
// stopSlaveWithMutex() is called during waiting mutex
|
|
return nil
|
|
}
|
|
mdb.replication.running.Add(1)
|
|
defer mdb.replication.running.Done()
|
|
for {
|
|
select {
|
|
case payload, open := <-mdb.replication.masterChan:
|
|
if !open {
|
|
return errors.New("master channel unexpected close")
|
|
}
|
|
if payload.Err != nil {
|
|
return payload.Err
|
|
}
|
|
cmdLine, ok := payload.Data.(*protocol.MultiBulkReply)
|
|
if !ok {
|
|
return errors.New("unexpected payload: " + string(payload.Data.ToBytes()))
|
|
}
|
|
mdb.replication.mutex.Lock()
|
|
if mdb.replication.modCount != modCount {
|
|
// replication conf changed during connecting and waiting mutex
|
|
return nil
|
|
}
|
|
mdb.Exec(conn, cmdLine.Args)
|
|
// todo: directly get size from socket
|
|
n := len(cmdLine.ToBytes())
|
|
mdb.replication.replOffset += int64(n)
|
|
logger.Info(fmt.Sprintf("receive %d bytes from master, current offset %d",
|
|
n, mdb.replication.replOffset))
|
|
mdb.replication.mutex.Unlock()
|
|
case <-done:
|
|
return nil
|
|
}
|
|
}
|
|
}
|
|
|
|
func (mdb *MultiDB) slaveCron() {
|
|
repl := mdb.replication
|
|
if repl.masterConn == nil {
|
|
return
|
|
}
|
|
replTimeout := 60 * time.Second
|
|
if config.Properties.ReplTimeout != 0 {
|
|
replTimeout = time.Duration(config.Properties.ReplTimeout) * time.Second
|
|
}
|
|
minLastRecvTime := time.Now().Add(-replTimeout)
|
|
if repl.lastRecvTime.Before(minLastRecvTime) {
|
|
// reconnect with master
|
|
err := mdb.reconnectWithMaster()
|
|
if err != nil {
|
|
logger.Error("send failed " + err.Error())
|
|
}
|
|
return
|
|
}
|
|
// send ack to master
|
|
err := repl.sendAck2Master()
|
|
if err != nil {
|
|
logger.Error("send failed " + err.Error())
|
|
}
|
|
}
|
|
|
|
// Send a REPLCONF ACK command to the master to inform it about the current processed offset
|
|
func (repl *replicationStatus) sendAck2Master() error {
|
|
psyncCmdLine := utils.ToCmdLine("REPLCONF", "ACK",
|
|
strconv.FormatInt(repl.replOffset, 10))
|
|
psyncReq := protocol.MakeMultiBulkReply(psyncCmdLine)
|
|
_, err := repl.masterConn.Write(psyncReq.ToBytes())
|
|
//logger.Info("send ack to master")
|
|
return err
|
|
}
|
|
|
|
func (mdb *MultiDB) reconnectWithMaster() error {
|
|
logger.Info("reconnecting with master")
|
|
mdb.replication.mutex.Lock()
|
|
defer mdb.replication.mutex.Unlock()
|
|
mdb.replication.stopSlaveWithMutex()
|
|
go mdb.syncWithMaster()
|
|
return nil
|
|
}
|