I have a Go program that generates a lot of HTTP requests from multiple goroutines. after running for a while, the program spits out an error: connect: cannot assign requested address.
When checking with netstat, I get a high number (28229) of connections in TIME_WAIT.
The high number of TIME_WAIT sockets happens when I the number of goroutines is 3 and is severe enough to cause a crash when it is 5.
I run Ubuntu 14.4 under docker and go version 1.7
This is the Go program.
package main
import (
"io/ioutil"
"log"
"net/http"
"sync"
)
var wg sync.WaitGroup
var url="http://172.17.0.9:3000/";
const num_coroutines=5;
const num_request_per_coroutine=100000
func get_page(){
response, err := http.Get(url)
if err != nil {
log.Fatal(err)
} else {
defer response.Body.Close()
_, err =ioutil.ReadAll(response.Body)
if err != nil {
log.Fatal(err)
}
}
}
func get_pages(){
defer wg.Done()
for i := 0; i < num_request_per_coroutine; i++{
get_page();
}
}
func main() {
for i:=0;i<num_coroutines;i++{
wg.Add(1)
go get_pages()
}
wg.Wait()
}
This is the server program:
package main
import (
"fmt"
"net/http"
"log"
)
var count int;
func sayhelloName(w http.ResponseWriter, r *http.Request) {
count++;
fmt.Fprintf(w,"Hello World, count is %d",count) // send data to client side
}
func main() {
http.HandleFunc("/", sayhelloName) // set router
err := http.ListenAndServe(":3000", nil) // set listen port
if err != nil {
log.Fatal("ListenAndServe: ", err)
}
}
The default http.Transport is opening and closing connections too quickly. Since all connections are to the same host:port combination, you need to increase MaxIdleConnsPerHost to match your value for num_coroutines. Otherwise, the transport will frequently close the extra connections, only to have them reopened immediately.
You can set this globally on the default transport:
http.DefaultTransport.(*http.Transport).MaxIdleConnsPerHost = numCoroutines
Or when creating your own transport
t := &http.Transport{
Proxy: http.ProxyFromEnvironment,
DialContext: (&net.Dialer{
Timeout: 30 * time.Second,
KeepAlive: 30 * time.Second,
}).DialContext,
MaxIdleConnsPerHost: numCoroutines,
MaxIdleConns: 100,
IdleConnTimeout: 90 * time.Second,
TLSHandshakeTimeout: 10 * time.Second,
ExpectContinueTimeout: 1 * time.Second,
}
Similar question: Go http.Get, concurrency, and "Connection reset by peer"
Related
I made a server with client hitting throught http. I set retry mechanism in the client inside its Transport's RoundTripper method. Here's the example of working code for each server and client:
server main.go
package main
import (
"fmt"
"net/http"
"time"
)
func test(w http.ResponseWriter, req *http.Request) {
time.Sleep(2 * time.Second)
fmt.Fprintf(w, "hello\n")
}
func main() {
http.HandleFunc("/test", test)
http.ListenAndServe(":8090", nil)
}
client main.go
package main
import (
"context"
"fmt"
"log"
"net/http"
"time"
)
type Retry struct {
nums int
transport http.RoundTripper
}
// to retry
func (r *Retry) RoundTrip(req *http.Request) (resp *http.Response, err error) {
for i := 0; i < r.nums; i++ {
log.Println("Attempt: ", i+1)
resp, err = r.transport.RoundTrip(req)
if resp != nil && err == nil {
return
}
log.Println("Retrying...")
}
return
}
func main() {
r := &Retry{
nums: 5,
transport: http.DefaultTransport,
}
c := &http.Client{Transport: r}
// each request will be timeout in 1 second
ctx, cancel := context.WithTimeout(context.Background(), 1*time.Second)
defer cancel()
req, err := http.NewRequestWithContext(ctx, http.MethodGet, "http://localhost:8090/test", nil)
if err != nil {
panic(err)
}
resp, err := c.Do(req)
if err != nil {
panic(err)
}
fmt.Println(resp.StatusCode)
}
What's happening is the retry seems only work for first iteration. For subsequent iteration it doesn't wait for one second each, instead the debugging message printed for as much as the retry nums.
I expect the retry attempt to be waiting 1 second each as I put the timeout for 1 second in the context. But it seems only wait for 1 second for whole retries. What do I miss?
Beside, how to stop server from processing timeout request?, I saw CloseNotifier already deprecated.
The problem is with the context. Once the context is done, you cannot reuse the same context anymore. You have to re-create the context at every attempt. You can get the timeout from parent context, and use it to create new context with it.
func (r *retry) RoundTrip(req *http.Request) (resp *http.Response, err error) {
var (
duration time.Duration
ctx context.Context
cancel func()
)
if deadline, ok := req.Context().Deadline(); ok {
duration = time.Until(deadline)
}
for i := 0; i < r.nums; i++ {
if duration > 0 {
ctx, cancel = context.WithTimeout(context.Background(), duration)
req = req.WithContext(ctx)
}
resp, err = r.rt.RoundTrip(req)
...
// the rest of code
...
}
return
}
This code will create new fresh context at every attempt by using the timeout from its parent.
For the server you can use the Request.Context() to check if a request is cancelled or not.
In the client the request times out when the context times out after 1 second. So the context does not trigger the roundtrip period. If you want the request to retry before the context is done you should change the behaviour of the transport you are using. You are now using the http.DefaultTransport which is defined as follows:
var DefaultTransport RoundTripper = &Transport{
Proxy: ProxyFromEnvironment,
DialContext: (&net.Dialer{
Timeout: 30 * time.Second,
KeepAlive: 30 * time.Second,
DualStack: true,
}).DialContext,
ForceAttemptHTTP2: true,
MaxIdleConns: 100,
IdleConnTimeout: 90 * time.Second,
TLSHandshakeTimeout: 10 * time.Second,
ExpectContinueTimeout: 1 * time.Second,
}
Transport has more time out variables these set here so depending on when you want to retry you should set the appropriate time out. For example in your case you could set Transport.ResponseHeaderTimeout to 1 second. When the server does not reply with a response header within 1 second the client will retry. You then make your context time out after 5 (or better 6) seconds you should see the client retrying the amount you specified (5 times).
I am working on a Go app that has a web server. I was trying to add timeouts and encountered an issue. Here's a sample code I made to reproduce it because posting the actual code would be impossible:
package main
import (
"fmt"
"html/template"
"net/http"
"time"
)
var layout *template.Template
func main() {
router := http.NewServeMux()
server := &http.Server{
Addr: ":8888",
Handler: router,
ReadTimeout: 5 * time.Second,
WriteTimeout: 1 * time.Second,
IdleTimeout: 15 * time.Second,
}
router.HandleFunc("/", home)
var err error
layout, err = template.ParseFiles("./layout.html")
if err != nil {
fmt.Printf("Error1: %+v\n", err)
}
server.ListenAndServe()
}
func home(w http.ResponseWriter, r *http.Request) {
fmt.Println("responding")
err := layout.Execute(w, template.HTML(`World`))
if err != nil {
fmt.Printf("Error2: %+v\n", err)
}
time.Sleep(5 * time.Second)
}
layout.html: Hello {{.}}!
When I run it and visit 127.0.0.1:8888, the browser stays loading, and the home() which is triggering the timeout, starts over again and it does it 10 times before it stops and the browser shows a connection reset error.
I was expecting that after a timeout, the func would end immediately, the connection be closed and the browser stop loading and show an error.
How can I achieve this?
immediately response use goroutines and context timeout
package main
import (
"context"
"fmt"
"html/template"
"net/http"
"time"
)
var layout *template.Template
var WriteTimeout = 1 * time.Second
func main() {
router := http.NewServeMux()
server := &http.Server{
Addr: ":8889",
Handler: router,
ReadTimeout: 5 * time.Second,
WriteTimeout: WriteTimeout + 10*time.Millisecond, //10ms Redundant time
IdleTimeout: 15 * time.Second,
}
router.HandleFunc("/", home)
server.ListenAndServe()
}
func home(w http.ResponseWriter, r *http.Request) {
fmt.Printf("responding\n")
ctx, _ := context.WithTimeout(context.Background(), WriteTimeout)
worker, cancel := context.WithCancel(context.Background())
var buffer string
go func() {
// do something
time.Sleep(2 * time.Second)
buffer = "ready all response\n"
//do another
time.Sleep(2 * time.Second)
cancel()
fmt.Printf("worker finish\n")
}()
select {
case <-ctx.Done():
//add more friendly tips
w.WriteHeader(http.StatusInternalServerError)
return
case <-worker.Done():
w.Write([]byte(buffer))
fmt.Printf("writed\n")
return
}
}
The http.Request struct includes the remote IP and port of the request's sender:
// RemoteAddr allows HTTP servers and other software to record
// the network address that sent the request, usually for
// logging. This field is not filled in by ReadRequest and
// has no defined format. The HTTP server in this package
// sets RemoteAddr to an "IP:port" address before invoking a
// handler.
// This field is ignored by the HTTP client.
**RemoteAddr string**
The http.Response object has no such field.
I would like to know the IP address that responded to the request I sent, even when I sent it to a DNS address.
I thought that net.LookupHost() might be helpful, but 1) it can return multiple IPs for a single host name, and 2) it ignores the hosts file unless cgo is available, which it is not in my case.
Is it possible to retrieve the remote IP address for an http.Response?
Use the net/http/httptrace package and use the GotConnInfo hook to capture the net.Conn and its corresponding Conn.RemoteAddr().
This will give you the address the Transport actually dialled, as opposed to what was resolved in DNSDoneInfo:
package main
import (
"log"
"net/http"
"net/http/httptrace"
)
func main() {
req, err := http.NewRequest("GET", "https://example.com/", nil)
if err != nil {
log.Fatal(err)
}
trace := &httptrace.ClientTrace{
GotConn: func(connInfo httptrace.GotConnInfo) {
log.Printf("resolved to: %s", connInfo.Conn.RemoteAddr())
},
}
req = req.WithContext(httptrace.WithClientTrace(req.Context(), trace))
client := &http.Client{}
_, err := client.Do(req)
if err != nil {
log.Fatal(err)
}
}
Outputs:
~ go run ip.go
2017/02/18 19:38:11 resolved to: 104.16.xx.xxx:443
Another solution I came up with was the hook the DialContext function in the http client transport. This is a specific solution that lets you modify the http.Client instead of the request which may be useful.
We first create a function that returns a hooked dial context
func remoteAddressDialHook(remoteAddressPtr *net.Addr) func(ctx context.Context, network string, address string) (net.Conn, error) {
hookedDialContext := func(ctx context.Context, network, address string) (net.Conn, error) {
originalDialer := &net.Dialer{
Timeout: 30 * time.Second,
KeepAlive: 30 * time.Second,
}
conn, err := originalDialer.DialContext(ctx, network, address)
if err != nil {
return nil, err
}
// conn was successfully created
*remoteAddressPtr = conn.RemoteAddr()
return conn, err
}
return hookedDialContext
}
We can then use this function to create a DialContext that writes to an outparameter
var remoteAddr net.Addr
customTransport := &http.Transport{
Proxy: http.ProxyFromEnvironment,
DialContext: remoteAddressDialHook(&remoteAddr),
ForceAttemptHTTP2: true,
MaxIdleConns: 100,
IdleConnTimeout: 90 * time.Second,
TLSHandshakeTimeout: 10 * time.Second,
ExpectContinueTimeout: 1 * time.Second,
}
customHttpClient := http.Client{
Transport: customTransport,
}
// do what you normally would with a http client, it will then set the remoteAddr to be the remote address
fmt.Println(remoteAddr.String())
I have a client machine with multiple NICs, how do I bind an http.Client in Go to a certain NIC or to a certain SRC IP Address?
Say you have some very basic http client code that looks like:
package main
import (
"net/http"
)
func main() {
webclient := &http.Client{}
req, _ := http.NewRequest("GET", "http://www.google.com", nil)
httpResponse, _ := webclient.Do(req)
defer httpResponse.Body.Close()
}
Is there a way to bind to a certain NIC or IP?
Similar to this question, you need to set the http.Client.Transport field. Setting it to an instance of net.Transport allows you to specify which net.Dialer you want to use. net.Dialer then allows you to specify the local address to make connections from.
Example:
localAddr, err := net.ResolveIPAddr("ip", "<my local address>")
if err != nil {
panic(err)
}
// You also need to do this to make it work and not give you a
// "mismatched local address type ip"
// This will make the ResolveIPAddr a TCPAddr without needing to
// say what SRC port number to use.
localTCPAddr := net.TCPAddr{
IP: localAddr.IP,
}
webclient := &http.Client{
Transport: &http.Transport{
Proxy: http.ProxyFromEnvironment,
DialContext: (&net.Dialer{
LocalAddr: &localTCPAddr,
Timeout: 30 * time.Second,
KeepAlive: 30 * time.Second,
DualStack: true,
}).DialContext,
MaxIdleConns: 100,
IdleConnTimeout: 90 * time.Second,
TLSHandshakeTimeout: 10 * time.Second,
ExpectContinueTimeout: 1 * time.Second,
},
}
Here is a fully working example that incorporates the answer from Tim. I also broke out all of the nested pieces to make it easier to read and learn from.
package main
import (
"fmt"
"io/ioutil"
"net"
"net/http"
"time"
)
func main() {
localAddr, err := net.ResolveIPAddr("ip", "10.128.64.219")
if err != nil {
panic(err)
}
localTCPAddr := net.TCPAddr{
IP: localAddr.IP,
}
d := net.Dialer{
LocalAddr: &localTCPAddr,
Timeout: 30 * time.Second,
KeepAlive: 30 * time.Second,
}
tr := &http.Transport{
Proxy: http.ProxyFromEnvironment,
Dial: d.Dial,
TLSHandshakeTimeout: 10 * time.Second,
}
webclient := &http.Client{Transport: tr}
// Use NewRequest so we can change the UserAgent string in the header
req, err := http.NewRequest("GET", "http://www.google.com:80", nil)
if err != nil {
panic(err)
}
res, err := webclient.Do(req)
if err != nil {
panic(err)
}
fmt.Println("DEBUG", res)
defer res.Body.Close()
content, err := ioutil.ReadAll(res.Body)
if err != nil {
panic(err)
}
fmt.Printf("%s", string(content))
}
What is the upper limit on the number of concurrent HTTP connections that a very simple server implemented in Go can handle?
The number of concurrent HTTP connections is limited by available memory and by operating system limits.
In Linux, the soft operating system limits - such as the maximum number of open files - can be printed out and changed by using ulimit.
In terms of memory, each HTTP connection in a minimal Go HTTP server running on 32-bit Linux consumes 21 KiB of memory (the source code of this server, compilable with Go version 2013-03-23, is below). On 64-bit Linux, the memory consumption can be expected to be higher.
On a 32-bit system with 1GB of memory available to the server, 21 KiB means that about 50,000 simultaneous connections are possible. This does not include memory consumed by the Linux kernel.
package main
import (
"flag"
"fmt"
"net/http"
"os"
"runtime"
"sync"
)
var isClient = flag.Bool("client", false, "Whether to start the HTTP server or the HTTP client")
var N = flag.Int("n", 1000, "Number of concurrent HTTP requests")
var wait = make(chan byte)
var counter = 0
var reachedN = make(chan byte)
func handler(w http.ResponseWriter, req *http.Request) {
fmt.Fprintf(w, "some text")
counter++
if counter == *N {
reachedN <- 0
}
<-wait // Block this goroutine
}
func main() {
flag.Parse()
if *N <= 0 {
fmt.Fprintf(os.Stderr, "invalid number of goroutines")
os.Exit(1)
}
if *isClient {
// Initiate N http connections
var wg sync.WaitGroup
for i := 0; i < *N; i++ {
wg.Add(1)
go func(ii int) {
_, err := http.Get("http://127.0.0.1:12345")
if err != nil {
fmt.Fprintf(os.Stderr, "client %d: %s\n", ii, err)
os.Exit(1)
}
wg.Done()
}(i)
}
wg.Wait()
} else {
runtime.GOMAXPROCS(1) // No concurrency
// Read MemStats
var m0 runtime.MemStats
runtime.ReadMemStats(&m0)
go func() {
<-reachedN // Wait until there are *N concurrent requests
// Read MemStats
var m1 runtime.MemStats
runtime.ReadMemStats(&m1)
fmt.Printf("Number of HTTP connections: %d\n", *N)
fmt.Printf("Memory consumption per connection: %.2f bytes\n", float64(m1.Sys-m0.Sys)/float64(*N))
os.Exit(1)
}()
http.HandleFunc("/", handler)
err := http.ListenAndServe(":12345", nil)
if err != nil {
fmt.Fprintf(os.Stderr, "server: %s\n", err)
os.Exit(1)
}
}
}