func (f *fileOutputer) extract(bp *bytes.Buffer) {
buf := make([]byte, 4)
bp.Read(buf)
l, _ := binary.Uvarint(buf)
headerLen := int(l)
//get pack header
buf = make([]byte, headerLen)
bp.Read(buf)
header := tcp_pack.ParseHeader(buf)
r, err := zlib.NewReader(bp)
if err != nil {
loglib.Error("zlib reader Error: " + err.Error())
} else {
lines, _ := strconv.Atoi(header["lines"])
done := false
if header["done"] == "1" {
done = true
}
f.ic.Add(header["ip"], header["hour"], header["id"], lines, done)
writerKey := header["ip"] + "_" + header["hour"]
fout := f.getWriter(f.writers, f.dataDir, writerKey)
//一头一尾写头信息，节省硬盘
buf = append(buf, '\n')
//fout.Write(buf)
nn, err := io.Copy(fout, r)
if err != nil {
loglib.Warning(fmt.Sprintf("save %s_%s_%s error:%s, saved:%d", header["ip"], header["hour"], header["id"], err, nn))
}
//fout.Write(buf)
//单独存一份header便于查数
fout = f.getWriter(f.headerWriters, f.headerDir, writerKey)
n, err := fout.Write(buf)
if err != nil {
loglib.Info(fmt.Sprintf("writer header %s %d %s", writerKey, n, err.Error()))
}
if done || time.Now().Unix() > f.checkTime.Unix() {
hourFinish, _ := f.ic.Check()
for ip, hours := range hourFinish {
for _, hour := range hours {
writerKey = ip + "_" + hour
}
}
f.closeWriters(f.writers)
f.closeWriters(f.headerWriters)
f.checkTime.Add(2 * time.Minute)
}
r.Close()
}
}
func (this *MongoDbOutputer) extract(bp *bytes.Buffer) (r io.ReadCloser, packId string, date string, err error) {
buf := make([]byte, 4)
bp.Read(buf)
l, _ := binary.Uvarint(buf)
headerLen := int(l)
//get pack header
buf = make([]byte, headerLen)
bp.Read(buf)
header := tcp_pack.ParseHeader(buf)
r, err = zlib.NewReader(bp)
if err != nil {
loglib.Error("zlib reader Error: " + err.Error())
}
date = header["hour"][0:8] //用于按天分库
packId = fmt.Sprintf("%s_%s_%s", header["ip"], header["hour"], header["id"])
return
}
func (t *TcpReceiver) handleConnnection(conn net.Conn, wg *sync.WaitGroup) {
defer func() {
if err := recover(); err != nil {
loglib.Error(fmt.Sprintf("tcp receiver connection panic:%v", err))
}
conn.Close()
wg.Done()
}()
/*
用于标识收到退出信号后，能否直接退出
只要接收信号时，包没有收完，都是可退出的，
发送方会缓存以后重传;
如果收完了就不能直接退出，可能包已传给下一级处理但是
却告诉发送方发送失败
*/
var quit = false //用于标识是否要退出
go lib.HandleQuitSignal(func() {
//关闭连接，避免阻塞在网络io上
conn.Close()
quit = true
})
request := make([]byte, 512*1024) //缓冲为512k
var packLen int = 0
currLen := 0
var b = new(bytes.Buffer)
var content = new(bytes.Buffer)
inAddr := conn.RemoteAddr().String()
parts := strings.Split(inAddr, ":")
inIp := parts[0]
packId := "unkown"
var routeInfo map[string]string
var rePull = false //是否补拉，如果是补拉就不做重复包检验
loglib.Info("incoming: " + inAddr)
outer:
for !quit {
st := time.Now()
if packLen == 0 {
conn.SetReadDeadline(time.Now().Add(5 * time.Minute))
time1 := time.Now() //时间打点
// read zlib pack header length
buf := make([]byte, 4)
_, err := conn.Read(buf)
if err != nil {
loglib.Warning(fmt.Sprintf("conn:%s, get header len, tcp receiver read error:%s, elapse:%s", inAddr, err.Error(), time.Now().Sub(time1)))
break
}
l, _ := binary.Uvarint(buf)
headerLen := int(l)
//get pack header
headerBuf := make([]byte, headerLen)
time2 := time.Now()
_, err = conn.Read(headerBuf)
if err != nil {
loglib.Warning(fmt.Sprintf("conn:%s, get header, tcp receiver read error:%s, elapse:%s", inAddr, err.Error(), time.Now().Sub(time2)))
break
}
//是否补拉
route0 := tcp_pack.ParseHeader(headerBuf)
if v, ok := route0["repull"]; ok && v == "1" {
rePull = true
} else {
rePull = false
}
buf = append(buf, headerBuf...)
header, _, err := tcp_pack.ExtractHeader(buf)
if err != nil {
loglib.Error("wrong format header " + string(headerBuf) + " " + err.Error())
conn.Write([]byte("wrong header"))
break
}
packId = tcp_pack.GetPackId(buf)
packLen = header.PackLen
currLen = 0
routeInfo = make(map[string]string)
b = new(bytes.Buffer)
content = new(bytes.Buffer)
loglib.Info(fmt.Sprintf("conn:%s, start receive pack %s, pack len:%d, header len:%d, header elapse:%s", inAddr, packId, packLen, headerLen, time.Now().Sub(time1)))
b.Write(buf)
routeInfo["ip"] = lib.GetIp()
routeInfo["stage"] = "tcp recv"
routeInfo["st"] = st.Format("2006-01-02 15:04:05.000")
}
//读包体的超时
conn.SetReadDeadline(time.Now().Add(5 * time.Minute))
time3 := time.Now()
//read enough bytes
for currLen < packLen {
requestLen, err := conn.Read(request)
if requestLen == 0 || err != nil {
//sender有重发机制，所以可丢弃
packLen = 0 //设为0以便读取新的包
ed := time.Now()
loglib.Warning(fmt.Sprintf("conn:%s, not full! ip:%s, packid:%s, received:%d, end recv:%s, elapse:%s, body elapse:%s, error:%s", inAddr, inIp, packId, currLen, ed, ed.Sub(st), ed.Sub(time3), err.Error()))
break outer //连接出错直接跳出外层循环
}
currLen += requestLen
content.Write(request[:requestLen])
}
if packLen > 0 && currLen >= packLen {
//收完马上应答
_, err := conn.Write([]byte("ok"))
if err != nil {
loglib.Warning(fmt.Sprintf("ip:%s, packid:%s received, but response back error:%s", inIp, packId, err.Error()))
} else {
loglib.Info(fmt.Sprintf("conn:%s, response to packid:%s", inAddr, packId))
}
//避免收到重复包（补拉例外）
appeared, ok, code := t.hasAppeared(content)
if !ok || rePull {
ed := time.Now()
routeInfo["ed"] = ed.Format("2006-01-02 15:04:05.000")
routeInfo["elapse"] = ed.Sub(st).String()
b.Write(content.Bytes())
vbytes := tcp_pack.Packing(b.Bytes(), routeInfo, true)
b = bytes.NewBuffer(vbytes)
t.buffer <- *b
packAppear := PackAppear{time.Now().Unix(), packId}
t.mutex.Lock()
t.footPrint[code] = packAppear //这里挂过
t.mutex.Unlock()
loglib.Info(fmt.Sprintf("conn:%s, finish ip:%s, packid:%s, repull:%v, received:%d, elapse:%s, body elapse:%s", inAddr, inIp, packId, rePull, currLen, ed.Sub(st), ed.Sub(time3)))
} else {
loglib.Info(fmt.Sprintf("conn:%s, pack %s repeat %s already appear at %s", inAddr, packId, appeared.Id, time.Unix(appeared.Time, 0)))
}
packLen = 0
}
}
loglib.Info("conn finish: " + inAddr)
}
func (e *etlOutputer) runEtl(spiderList string, colsFile string, hostsList string, ipBlackList string) {
wg := &sync.WaitGroup{}
fkeyChan := make(chan string, 100)
defer func() {
if err := recover(); err != nil {
loglib.Error(fmt.Sprintf("runEtl() panic:%v", err))
}
e.ic.SaveStatus()
e.closeWriters(e.writers)
e.closeWriters(e.headerWriters)
close(fkeyChan)
//等待etl routine结束
wg.Wait()
}()
for i := 0; i < 5; i++ {
wg.Add(1)
go e.doEtl(fkeyChan, e.dataDir, e.etlDir, e.etlDoneDir, e.etlFailDir, spiderList, colsFile, hostsList, ipBlackList, wg)
}
nextCheckTime := time.Now().Add(2 * time.Minute)
//使用range遍历，方便安全退出，只要发送方退出时关闭chan，这里就可以退出了
for b := range e.buffer {
loglib.Info(fmt.Sprintf("pack in chan: %d", len(e.buffer)))
buf := make([]byte, 4)
bp := &b
bp.Read(buf)
l, _ := binary.Uvarint(buf)
headerLen := int(l)
//get pack header
buf = make([]byte, headerLen)
bp.Read(buf)
header := tcp_pack.ParseHeader(buf)
r, err := zlib.NewReader(bp)
if err != nil {
loglib.Error("zlib reader Error: " + err.Error())
} else {
lines, _ := strconv.Atoi(header["lines"])
done := false
if header["done"] == "1" {
done = true
}
e.ic.Add(header["ip"], header["hour"], header["id"], lines, done)
writerKey := header["ip"] + "_" + header["hour"]
fout := e.getWriter(e.writers, e.dataDir, writerKey)
buf = append(buf, '\n')
/*
//一头一尾写头信息，节省硬盘
n, err := fout.Write(buf)
if err != nil {
loglib.Info(fmt.Sprintf("write %s %d %s", writerKey, n, err.Error()))
}
*/
nn, err := io.Copy(fout, r)
if err != nil {
loglib.Warning(fmt.Sprintf("save %s_%s_%s error:%s, saved:%d", header["ip"], header["hour"], header["id"], err, nn))
}
//fout.Write(buf)
//单独存一份header便于查数
fout = e.getWriter(e.headerWriters, e.headerDir, writerKey)
n, err := fout.Write(buf)
if err != nil {
loglib.Info(fmt.Sprintf("writer header %s %d %s", writerKey, n, err.Error()))
}
//增加2分钟check一次的规则，避免done包先到，其他的包未到，则可能要等到下一小时才能check
if done || time.Now().Unix() > nextCheckTime.Unix() {
hourFinish, _ := e.ic.Check()
for ip, hours := range hourFinish {
for _, hour := range hours {
writerKey = ip + "_" + hour
loglib.Info(fmt.Sprintf("fkeychan %d", len(fkeyChan)))
fkeyChan <- writerKey
}
}
e.closeWriters(e.writers)
e.closeWriters(e.headerWriters)
e.ic.SaveStatus()
nextCheckTime = time.Now().Add(2 * time.Minute)
}
r.Close()
}
}
}