In my project, I need send http requests 5000 times/s. I did read some blogs and questions about the problem, and I try things:
content.BodyBut they just don't work. Indeed, my app can run for several minutes or hours, then it reach to the ulimit value in my settings(first 50000, late 100000, use shell cmd ls -l /proc/PID/fd | wc -l). The http request is as follows:
package http
import "downloader/request"
import "downloader/response"
import "downloader/proxy"
import "downloader/ssdb"
import "net/http"
import "net/url"
import "time"
import "fmt"
import "strings"
import "io/ioutil"
import "errors"
func getProxy(ssdbClients []*ssdb.SSDBClient, proxyName string, reqUrl string) (string, error){
var err error
p := ""
url := ""
if proxyName != "" {
p, err = proxy.GetOne(ssdbClients,proxyName)
if err != nil {
return url, err
}
if strings.HasPrefix(reqUrl, "https") {
url = fmt.Sprintf("https://%s", p)
} else if strings.HasPrefix(reqUrl, "http") {
url = fmt.Sprintf("http://%s", p)
} else {
return url,errors.New(fmt.Sprintf("reqUrl %s not valid", reqUrl))
}
return url, err
}
return url, err
}
func customProxy(req *http.Request) (*url.URL, error) {
httpProxy := req.Header.Get("HttpProxy")
if httpProxy == "" {
return nil, nil
} else {
return url.Parse(httpProxy)
}
}
func GetClient() *http.Client {
var client http.Client
client = http.Client{
Timeout: 15 * time.Second,
}
transport := http.Transport{
Proxy: customProxy,
DisableCompression: false,
MaxIdleConnsPerHost: 10000,
}
client.Transport = &transport
return &client
}
func structResponse(req *request.HttpRequest, content *http.Response, proxy string) (response.HttpResponse, error) {
var resp response.HttpResponse
data, err := ioutil.ReadAll(content.Body)
if err != nil {
return resp, err
}
header := make(map[string]interface{})
for k, v := range content.Header {
if len(v) > 0 {
header[k] = v[0]
}
}
var encoding string
encoding = content.Header.Get("Content-Encoding")
resp = response.HttpResponse {
Request: *req,
ErrorCode: 0,
ErrorMsg: "",
StatusCode: content.StatusCode,
Reason: content.Status,
Html: string(data),
Headers: header,
Encoding: encoding,
Url: req.Url,
CrawlerName: req.CrawlerName,
ProxyName: req.ProxyName,
HttpProxy: proxy,
}
return resp, err
}
func Send(ssdbClients []*ssdb.SSDBClient, r *request.HttpRequest, httpClient *http.Client) (response.HttpResponse, error) {
var resp response.HttpResponse
var err error
var httpProxy string
httpProxy, err = getProxy(ssdbClients, r.ProxyName, r.Url)
if err != nil {
return resp, err
}
req, err := http.NewRequest(r.Method, r.Url, strings.NewReader(r.Data))
if err != nil {
return resp, err
}
u, err := url.Parse(r.Url)
if err != nil {
return resp, err
}
defaultHeaders := map[string] string {
"User-Agent": `Mozilla/5.0 (Macintosh; Intel Mac OS X 10_11_1)
AppleWebKit/537.36 (KHTML, like Gecko) Chrome/48.0.2564.109 Safari/537.36`,
"Upgrade-Insecure-Requests": "1",
"Connection": "keep-alive",
"Accept": "text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8",
"Accept-Language": "zh-CN,zh;q=0.8,en;q=0.6",
"Cache-Control": "max-age=0",
"Host": u.Host,
"HttpProxy": httpProxy,
}
for k, v := range defaultHeaders {
req.Header.Set(k, v)
}
for k, v := range r.Headers {
req.Header.Set(k, fmt.Sprintf("%v", v))
}
for k, v := range r.Cookies {
req.Header.Set("Cookie", fmt.Sprintf("%s=%v", k, v))
}
values := req.URL.Query()
for k, v := range r.Params {
values.Add(k, fmt.Sprintf("%v", v))
}
req.URL.RawQuery = values.Encode()
content, err := httpClient.Do(req)
if err != nil {
return resp, err
}
defer content.Body.Close()
resp, err = structResponse(r, content, httpProxy)
if err != nil {
return resp, err
}
respCookies := make(map[string] interface{})
if httpClient.Jar != nil {
cookies := httpClient.Jar.Cookies(req.URL)
for _, cookie := range cookies {
respCookies[cookie.Name] = cookie.Value
}
resp.Cookies = respCookies
}
return resp, err
}
And I use my http library as follows:
func sendRequest(clients []*ssdb.SSDBClient, r *request.HttpRequest, httpClient *http.Client) {
for {
resp, err := http.Send(clients, *r, httpClient)
if err != nil {
utils.Error.Println("http send fail ", err.Error())
} else {
if resp.StatusCode == 200 {
utils.Info.Println("status 200, success")
} else {
utils.Warning.Println("status ", resp.StatusCode, resp.Reason)
}
}
}
}
func main() {
httpClient := http.GetClient()
var wg sync.WaitGroup
wg.Add(1)
for i:=0;i<crawlerConsumers;i++ {
go sendRequest(clients, r, httpClient)
}
wg.Wait()
}
Can anyone have any ideas or suggestions?
Add Some profile info with pprof
/debug/pprof/
profiles:
0 block
99490 goroutine(This number is growing constantly, very strange)
1640 heap
14 threadcreate
# runtime.MemStats
# Alloc = 210967776
# TotalAlloc = 3673382376
# Sys = 646805032
# Lookups = 29226
# Mallocs = 15590422
# Frees = 14236863
# HeapAlloc = 210967776
# HeapSys = 331415552
# HeapIdle = 72081408
# HeapInuse = 259334144
# HeapReleased = 0
# HeapObjects = 1353559
# Stack = 283049984 / 283049984
# MSpan = 4007136 / 4112384
# MCache = 9664 / 16384
# BuckHashSys = 1575958
# NextGC = 306292818
# PauseNs = [2096153 2554867 2962532 4711468 3373778 3548443 1764893 1961992 2434108 2658626 2627832 2564733 2610212 3278169 2954882 2847604 4443650 3582780 3557718 4201288 4762243 8349689 3812924 5098353 5196422 4494087 5209715 5978150 5060982 4825367 6529020 5738726 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
# NumGC = 32
# EnableGC = true
# DebugGC = false
# runtime.MemStats
# Alloc = 334390488
# TotalAlloc = 7408124464
# Sys = 963748912
# Lookups = 64289
# Mallocs = 31759565
# Frees = 29798637
# HeapAlloc = 334390488
# HeapSys = 536903680
# HeapIdle = 171114496
# HeapInuse = 365789184
# HeapReleased = 0
# HeapObjects = 1960928
# Stack = 381648896 / 381648896
# MSpan = 4958240 / 5160960
# MCache = 9664 / 16384
# BuckHashSys = 1625694
# NextGC = 337299393
# PauseNs = [2096153 2554867 2962532 4711468 3373778 3548443 1764893 1961992 2434108 2658626 2627832 2564733 2610212 3278169 2954882 2847604 4443650 3582780 3557718 4201288 4762243 8349689 3812924 5098353 5196422 4494087 5209715 5978150 5060982 4825367 6529020 5738726 5563959 6997601 6906020 6778559 7592484 7193865 7198439 7815078 7217336 6923856 7127406 7491410 7029097 7757883 6861948 7295746 7245947 9037505 7656431 7322897 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
# NumGC = 52
# EnableGC = true
# DebugGC = false
# runtime.MemStats
# Alloc = 468642296
# TotalAlloc = 25479911160
# Sys = 1185490448
# Lookups = 278984
# Mallocs = 119997794
# Frees = 117189156
# HeapAlloc = 468642296
# HeapSys = 680919040
# HeapIdle = 174424064
# HeapInuse = 506494976
# HeapReleased = 0
# HeapObjects = 2808638
# Stack = 448397312 / 448397312
# MSpan = 6943104 / 7061504
# MCache = 9664 / 16384
# BuckHashSys = 1757422
# NextGC = 485127068
# PauseNs = [2096153 2554867 2962532 4711468 3373778 3548443 1764893 1961992 2434108 2658626 2627832 2564733 2610212 3278169 2954882 2847604 4443650 3582780 3557718 4201288 4762243 8349689 3812924 5098353 5196422 4494087 5209715 5978150 5060982 4825367 6529020 5738726 5563959 6997601 6906020 6778559 7592484 7193865 7198439 7815078 7217336 6923856 7127406 7491410 7029097 7757883 6861948 7295746 7245947 9037505 7656431 7322897 7155298 7288230 8027051 7442561 7528532 8418744 7847097 8791562 7242055 8418159 7677703 8662475 8408315 8562446 8110168 7308570 8593523 7820707 8734530 8981041 8721316 8152324 8390552 8603397 9059668 8153933 7986519 7744819 8491656 8562181 7816543 9145512 7902742 7780778 7636659 13573779 12684501 94808535 9443609 8257986 9072718 9634563 9229626 9449536 8644605 8898286 8499036 13101964 8743251 9119720 9267487 8178551 8444107 8362330 8447271 8558115 8788773 9977627 9058283 8743149 8508649 9099904 9382220 9684119 9789404 9730475 8506223 10363233 9979499 9033233 9634088 9739395 9129433 9122154 8615491 9632523 9314836 9845890 9888849 10034358 10207887 10073912 9892683 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
# NumGC = 135
# EnableGC = true
# DebugGC = false
Add cpu profile .svg file attachment
Finally, I figure it out that I should add timeout to the http client as follows:
transport := http.Transport{
Proxy: customProxy,
DisableCompression: false,
MaxIdleConnsPerHost: 10000,
Dial: (&net.Dialer{
Timeout: 15 * time.Second,
KeepAlive: 15 * time.Second,
}).Dial,
TLSHandshakeTimeout: 5 * time.Second,
}
have you looked at this yet: http://craigwickesser.com/2015/01/golang-http-to-many-open-files/
basically he suggests using req.Header.Set("Connection", "close")
I wonder if you're connections are staying open until the timeout/keepalive is hit, so manually setting it to less than the default is closing the connection fast enough where you don't overflow?