func clearDb(dbName string, swarm *swarm.Swarm) error {
cmd := []string{}
cmd = append(cmd, "mysql")
cmd = append(cmd, "-u"+"cup_dba")
cmd = append(cmd, "-p"+"111111")
cmd = append(cmd, "-S"+"/DBAASDAT/upsql.sock")
cmd = append(cmd, "-e drop database qjrtest1")
res, err := swarm.Exec(cmd, dbName)
if err != nil {
log.WithFields(log.Fields{
"DbName": dbName,
"err": err.Error(),
}).Error("drop qjrtest1 fail")
return err
}
res.Close()
cmd = []string{}
cmd = append(cmd, "mysql")
cmd = append(cmd, "-u"+"cup_dba")
cmd = append(cmd, "-p"+"111111")
cmd = append(cmd, "-S"+"/DBAASDAT/upsql.sock")
cmd = append(cmd, "-e drop database qjrtest2")
res1, err := swarm.Exec(cmd, dbName)
if err != nil {
log.WithFields(log.Fields{
"DbName": dbName,
"err": err.Error(),
}).Error("drop qjrtest2 fail")
return err
}
res1.Close()
return nil
}
func clearMaster(dbName string, swm *swarm.Swarm) error {
err := swm.MasterReset(dbName)
if err != nil {
return err
}
return nil
}
func voteSl(swm *swarm.Swarm, topology *structs.Topology) (string, error) {
maxMasterLogFileNum := 0
maxReadMasterLogPos := 0
ch := make(chan []interface{})
maxDbName := ""
for k, v := range topology.DataNodeGroup["default"] {
if v.Type == consts.Slave && v.Status == consts.Normal {
dbName := k
go func() {
masterLogFileNum, readMasterLogPos := swm.PositionGet(dbName)
log.WithFields(log.Fields{
"SlaveName": dbName,
"MasterLogFileNum": masterLogFileNum,
"ReadMasterLogPos": readMasterLogPos,
}).Debug("Sl vote")
pos := []interface{}{dbName, masterLogFileNum, readMasterLogPos}
ch <- pos
}()
}
}
for i := 0; i < topology.DataNodeGroupNormalCount["default"]-2; i++ {
pos := <-ch
dbName := pos[0].(string)
masterLogFileNum := pos[1].(int)
readMasterLogPos := pos[2].(int)
if masterLogFileNum > maxMasterLogFileNum {
if readMasterLogPos > maxReadMasterLogPos {
maxMasterLogFileNum = masterLogFileNum
maxReadMasterLogPos = readMasterLogPos
maxDbName = dbName
}
}
}
// all sl unavaliable
if maxDbName == "" {
return "", errors.New(
"when isolate M|Sb, all slave are unavilable for vote, isolate failed and return to hope health check enter sl auto isolate")
}
maxDbInfo := topology.DataNodeGroup["default"][maxDbName]
log.WithFields(log.Fields{
"winer": maxDbName + "(" + maxDbInfo.Ip + ":" + strconv.Itoa(maxDbInfo.Port) + ")",
}).Debug("Sl vote")
return maxDbName, nil
}
func createDb(dbName, dbString string, swarm *swarm.Swarm) error {
cmd := []string{}
cmd = append(cmd, "mysql")
cmd = append(cmd, "-u"+"cup_dba")
cmd = append(cmd, "-p"+"111111")
cmd = append(cmd, "-S"+"/DBAASDAT/upsql.sock")
cmd = append(cmd, "-e create database "+dbString)
res, err := swarm.Exec(cmd, dbName)
if err != nil {
log.WithFields(log.Fields{
"DbName": dbName,
"err": err.Error(),
"Database": dbString,
}).Error("create test database fail")
return err
}
res.Close()
return nil
}
func clearSlave(dbName string, swm *swarm.Swarm) error {
slaveStatus, err := swm.SlaveShowStatus(dbName)
if err != nil {
return err
}
if slaveStatus.MasterIp == "" {
return nil
}
err = swm.SlaveStop(dbName)
if err != nil {
return err
}
err = swm.SlaveReset(dbName)
if err != nil {
return err
}
return nil
}
func changeMaster(swm *swarm.Swarm, fromName, toIp string, toPort int) error {
slaveStatus, err := swm.SlaveShowStatus(fromName)
if err != nil {
return err
}
log.WithFields(log.Fields{
"DbName": fromName,
"SlaveStatus": slaveStatus,
}).Debug("change master")
if slaveStatus.MasterIp == toIp && slaveStatus.MasterPort == toPort {
if slaveStatus.SlaveIoRunning == "No" && slaveStatus.SlaveSqlRunning == "No" {
// 1.4 start slave
err = swm.SlaveStart(fromName)
if err != nil {
return err
}
}
return nil
}
if slaveStatus.SlaveIoRunning == "Yes" || slaveStatus.SlaveSqlRunning == "Yes" {
// 1.1 stop slave
err = swm.SlaveStop(fromName)
if err != nil {
return err
}
}
if slaveStatus.MasterIp != "" {
// 1.2 reset slave
err = swm.SlaveReset(fromName)
if err != nil {
return err
}
}
// 1.3 get gtid_executed
gtidExecuted, err := swm.GetGtidExecuted(fromName)
if err != nil {
return err
}
if gtidExecuted != "" {
// 1.4 reset master
err = swm.MasterReset(fromName)
if err != nil {
return err
}
// 1.5 set gtid_purged
err = swm.SetGtidPurged(fromName, gtidExecuted)
if err != nil {
return err
}
}
// 1.6 change master
err = swm.ChangeMaster(fromName, toIp, toPort)
if err != nil {
return err
}
// 1.7 start slave
err = swm.SlaveStart(fromName)
if err != nil {
return err
}
return nil
}
// no need RLock
func RecoverDb(dbName string, rwMutex *sync.RWMutex, topology *structs.Topology, swm *swarm.Swarm) (error, []string) {
var err error
var steps []string
defer func() {
if err != nil {
log.WithFields(log.Fields{
"DbName": dbName,
"err": err.Error(),
}).Error("recover fail")
} else {
log.WithFields(log.Fields{
"DbName": dbName,
}).Info("recover success")
}
}()
var recoverDbInfo *structs.DatabaseInfo
for k, v := range topology.DataNodeGroup["default"] {
if k == dbName {
recoverDbInfo = v
break
}
}
if recoverDbInfo == nil {
return errors.New("recover db node not found"), steps
}
// 1. check m slave status
if recoverDbInfo.Type == consts.StandBy || recoverDbInfo.Type == consts.Slave {
mName, _ := getMaster(topology)
var masterSlStatus structs.SlaveStatus
masterSlStatus, err = swm.SlaveShowStatus(mName)
if err != nil {
steps = append(steps, fmt.Sprintf("Step1. Clear M status fail: %s\n", err.Error()))
return err, steps
}
if masterSlStatus.MasterIp != "" {
// 1.1 stop m slave
err = swm.SlaveStop(mName)
if err != nil {
steps = append(steps, fmt.Sprintf("Step1. Clear M status fail: %s\n", err.Error()))
return err, steps
}
// 1.2 reset m slave
err = swm.SlaveReset(mName)
if err != nil {
steps = append(steps, fmt.Sprintf("Step1. Clear M status fail: %s\n", err.Error()))
return err, steps
}
}
}
steps = append(steps, "Step1. Clear M status success\n")
// 2. recover change master
if recoverDbInfo.Type == consts.StandBy {
mName, mDbInfo := getMaster(topology)
err = changeMaster(swm, dbName, mDbInfo.Ip, mDbInfo.Port)
if err != nil {
log.WithFields(log.Fields{
"Sb": dbName + "(" + recoverDbInfo.Ip + ":" + strconv.Itoa(recoverDbInfo.Port) + ")",
"M": mName + "(" + mDbInfo.Ip + ":" + strconv.Itoa(mDbInfo.Port) + ")",
"err:": err.Error(),
}).Error("change master fail, return")
steps = append(steps, fmt.Sprintf("Step2. Sb change master to M fail: %s\n", err.Error()))
return err, steps
}
log.WithFields(log.Fields{
"Sb": dbName + "(" + recoverDbInfo.Ip + ":" + strconv.Itoa(recoverDbInfo.Port) + ")",
"M": mName + "(" + mDbInfo.Ip + ":" + strconv.Itoa(mDbInfo.Port) + ")",
}).Debug("change master success")
steps = append(steps, "Step2. Sb change master to M success\n")
} else if recoverDbInfo.Type == consts.Slave {
sbName, sbDbInfo := getStandBy(topology)
err = changeMaster(swm, dbName, sbDbInfo.Ip, sbDbInfo.Port)
if err != nil {
log.WithFields(log.Fields{
"Sl": dbName + "(" + recoverDbInfo.Ip + ":" + strconv.Itoa(recoverDbInfo.Port) + ")",
"Sb": sbName + "(" + sbDbInfo.Ip + ":" + strconv.Itoa(sbDbInfo.Port) + ")",
"err:": err.Error(),
}).Error("change master fail, return")
steps = append(steps, fmt.Sprintf("Step2. Sl change master to Sb fail: %s\n", err.Error()))
return err, steps
}
log.WithFields(log.Fields{
"Sl": dbName + "(" + recoverDbInfo.Ip + ":" + strconv.Itoa(recoverDbInfo.Port) + ")",
"Sb": sbName + "(" + sbDbInfo.Ip + ":" + strconv.Itoa(sbDbInfo.Port) + ")",
}).Info("change master success")
steps = append(steps, "Step2. Sl change master to Sb success\n")
}
// 3. update topology
rwMutex.Lock()
dbInfo := getDbInfo(dbName, topology)
dbInfo.Status = consts.Normal
updateVersionAndNormalCount(topology, 1)
rwMutex.Unlock()
return nil, steps
}
// no need RLock
func IsolateDB(dbName string, rwMutex *sync.RWMutex, topology *structs.Topology, swm *swarm.Swarm) (error, []string) {
var err error
steps := []string{}
defer func() {
if err != nil && err.Error() != "" {
log.WithFields(log.Fields{
"DbName": dbName,
"err": err.Error(),
}).Error("isolate fail")
} else {
log.WithFields(log.Fields{
"DbName": dbName,
}).Info("isolate success")
}
}()
var isolateType string
for k, v := range topology.DataNodeGroup["default"] {
if k == dbName {
isolateType = v.Type
break
}
}
if isolateType == "" {
err = errors.New("isolate db node not found")
return err, steps
}
btopology, _ := json.MarshalIndent(topology, "", " ")
log.WithFields(log.Fields{
"IsolateDbType": isolateType,
"DbName": dbName,
"topology": string(btopology),
}).Debug("isolate db")
// only m normal ||
// isolate sb && only m sb normal ||
// isolate sl
// no need swarm, update topology and return
if topology.DataNodeGroupNormalCount["default"] == 1 ||
(isolateType == consts.StandBy && topology.DataNodeGroupNormalCount["default"] == 2) ||
isolateType == consts.Slave {
rwMutex.Lock()
topology.DataNodeGroup["default"][dbName].Status = consts.Abnormal
updateVersionAndNormalCount(topology, -1)
rwMutex.Unlock()
steps = append(steps, fmt.Sprintf("Step1. Isolate %s success\n", dbName))
return nil, steps
}
if isolateType == consts.Master {
// 1.sb checkGtidDiff
sbName, sbDbInfo := getStandBy(topology)
err = swm.IsDelay(sbName)
if err != nil {
log.WithFields(log.Fields{
"DbName": sbName,
"err": err.Error(),
}).Warn("Sb delay check fail, continue")
err = nil
}
// 1. end ********************
// m sb normal
// update topology and return
if topology.DataNodeGroupNormalCount["default"] == 2 {
// 3.update topology
_, mDbInfo := getMaster(topology)
rwMutex.Lock()
sbDbInfo.Type = consts.Master // sb -> m
mDbInfo.Type = consts.StandBy // m -> sb
mDbInfo.Status = consts.Abnormal
updateVersionAndNormalCount(topology, -1)
steps = append(steps, fmt.Sprintf("Step1. Isolate %s success\n", dbName))
rwMutex.Unlock()
return nil, steps
// ********************
}
// has normal sl
// 2.vote sb
var maxGtidDbName string
maxGtidDbName, err = voteSl(swm, topology)
if err != nil {
steps = append(steps, fmt.Sprintf("Step2. Vote new Sb fail: %s\n", err.Error()))
return err, steps
}
steps = append(steps, fmt.Sprintf("Step2. Vote new Sb success:%s\n", maxGtidDbName))
// 2. end ********************
// 3.update topology
rwMutex.Lock()
mName, mDbInfo := getMaster(topology)
mDbInfo.Type = consts.Slave // m -> sl
mDbInfo.Status = consts.Abnormal
sbDbInfo.Type = consts.Master // sb -> master
slDbInfo := topology.DataNodeGroup["default"][maxGtidDbName]
slDbInfo.Type = consts.StandBy // maxGtidSl -> sb
updateVersionAndNormalCount(topology, -1)
rwMutex.Unlock()
// 3. end ********************
// 4.for other sl(except origin m) changeMaster to newSb
errs := otherSlChangeMasterParallel(topology, mName, maxGtidDbName, slDbInfo, swm)
if len(errs) != 0 {
steps = append(steps, fmt.Sprintf("Step3. Other Sl change master to new Sb fail:%v\n", errs))
} else {
steps = append(steps, "Step3. Other Sl change master to new Sb success\n")
}
// 4. end ********************
} else if isolateType == consts.StandBy {
sbName, sbDbInfo := getStandBy(topology)
// has normal sl
// 1.vote sb
var maxGtidDbName string
maxGtidDbName, err = voteSl(swm, topology)
if err != nil {
return err, steps
}
// 1. end ********************
// no need check gtidDiff
// 2. newsb change master to master
mName, mDbInfo := getMaster(topology)
slDbInfo := topology.DataNodeGroup["default"][maxGtidDbName]
err = changeMaster(swm, maxGtidDbName, mDbInfo.Ip, mDbInfo.Port)
if err != nil {
log.WithFields(log.Fields{
"Sb": maxGtidDbName + ":" + slDbInfo.Ip + strconv.Itoa(slDbInfo.Port),
"M": mName + ":" + mDbInfo.Ip + strconv.Itoa(mDbInfo.Port),
"err:": err.Error(),
}).Error("new Sb change master to M fail, return")
steps = append(steps, fmt.Sprintf("Step1. New Sb change master to M fail:%s\n", err.Error()))
return err, steps
}
log.WithFields(log.Fields{
"Sb": maxGtidDbName + ":" + slDbInfo.Ip + strconv.Itoa(slDbInfo.Port),
"M": mName + ":" + mDbInfo.Ip + strconv.Itoa(mDbInfo.Port),
"IsolateType": isolateType,
}).Debug("new Sb change master to M success")
steps = append(steps, "Step1. New Sb change master to M success\n")
// 2. end ********************
// 3.update topology
rwMutex.Lock()
sbDbInfo.Type = consts.Slave // sb -> sl
sbDbInfo.Status = consts.Abnormal
slDbInfo.Type = consts.StandBy // maxGtidSl -> sb
updateVersionAndNormalCount(topology, -1)
rwMutex.Unlock()
// 3. end ********************
// 4.for other sl(except origin sb) changeMaster to newSb
errs := otherSlChangeMasterParallel(topology, sbName, maxGtidDbName, slDbInfo, swm)
if len(errs) != 0 {
steps = append(steps, fmt.Sprintf("Step2. Other Sl change master to new Sb fail:%v\n", errs))
} else {
steps = append(steps, "Step2. Other Sl change master to new Sb success\n")
}
// 4. end ********************
} else {
return errors.New("Unhandle isolate logic"), steps
}
return nil, steps
}
func checkTopo(swarm *swarm.Swarm, topology *structs.Topology, mCheck bool, t *testing.T) error {
log.Println("***CheckTopo start***")
var mName, mIp, sbName, sbIp string
var mPort, sbPort int
for k, v := range topology.DataNodeGroup["default"] {
if v.Type == consts.Master && v.Status == consts.Normal {
mName = k
mIp = v.Ip
mPort = v.Port
} else if v.Type == consts.StandBy && v.Status == consts.Normal {
sbName = k
sbIp = v.Ip
sbPort = v.Port
}
}
if mCheck {
// check m
mSlaveStatus, err := swarm.SlaveShowStatus(mName)
if err != nil {
return fmt.Errorf(err.Error())
}
if mSlaveStatus.MasterIp != "" ||
mSlaveStatus.MasterPort != 0 {
return fmt.Errorf("master %s init/isolate/recover err, slave status:%#v\n", mName, mSlaveStatus)
}
}
if sbName == "" {
log.Println("***CheckTopo end***")
return nil
}
// check sb
sbSlaveStatus, err := swarm.SlaveShowStatus(sbName)
if err != nil {
return fmt.Errorf(err.Error())
}
if sbSlaveStatus.MasterIp != mIp ||
sbSlaveStatus.MasterPort != mPort ||
sbSlaveStatus.SlaveIoRunning != "Yes" ||
sbSlaveStatus.SlaveSqlRunning != "Yes" {
return fmt.Errorf("standby %s init/isolate/recover err, slave status:%#v\n", sbName, sbSlaveStatus)
}
// check sl
if topology.DataNodeGroupNormalCount["default"] > 2 {
for k, v := range topology.DataNodeGroup["default"] {
if v.Type == consts.Slave && v.Status == consts.Normal {
slSlaveStatus, err := swarm.SlaveShowStatus(k)
if err != nil {
return fmt.Errorf(err.Error())
}
if slSlaveStatus.MasterIp != sbIp ||
slSlaveStatus.MasterPort != sbPort ||
slSlaveStatus.SlaveIoRunning != "Yes" ||
slSlaveStatus.SlaveSqlRunning != "Yes" {
return fmt.Errorf("slave %s init/isolate/recover err, slave status:%#v\n", k, sbSlaveStatus)
}
}
}
}
log.Println("***CheckTopo end***")
return nil
}
