High Average Request Duration on Digitalocean Load balancer using golang fasthttp - go

I have a go program which takes requests processes them within 100ms and sends a response. I am using digitalocean Load Balancer and i noticed 'Average Request Duration' is pretty high around 33Kms.Below is the graph
https://i.stack.imgur.com/Uj9bN.png
Here is the code i am using...
count :="one"
n := func(ctx *fasthttp.RequestCtx) {
switch count {
case "one":
handle(ctx) //...dosomething for 100 ms function
default:
ctx.Error("not found", fasthttp.StatusNotFound)
}
}
fasthttp.ListenAndServe(":8082", n)
What am i doing wrong and why isn't my program closing requests.

Related

Golang API giving higher response time with increasing number of concurrent users

I am having some problems with the concurrent HTTP connection in the golang. Kindly read the whole question, and as the actual code is quite long, I am using pseudocode
In short, I have to create a single API, which will internally call 5 other APIs, unify their response, and send them as a single response.
I am using goroutines to call those 5 internal APIs along with timeout, and using channels to ensure that every goroutine has been completed, then I unify their response, and return the same.
Things are going fine when I do local testing, my response time is around 300ms, which is pretty good.
The problem arises when I do the locust load testing of 200 users, then my response time go as high as 7 8 sec. I am thinking it has to do with the HTTP client waiting for the resources as we are running a high number of goroutines.
like 1 API spin up 5 go-routine, so if each of 200 users makes API requests at the rate of supposing 5 req/sec. Then a total number of goroutines goes way higher. Again this is my assumption only
p.s. normally the API I am building is pretty good in response time,
I am using all the caching and stuff and any response greater than
400ms should not be the case
So can anyone please tell me how can I tackle this problem of
increasing response time when number of concurrent users increases
Locust test report
pseudo code
simple route
group.POST("/test", controller.testHandler)
controller
type Worker struct {
NumWorker int
Data chan structures.Placement
}
e := Worker{
NumWorker: 5, // Number of worker goroutine(s)
Data: make(chan, 5) /* Buffer Size */),
}
//call the goroutines along with the
for i := 0; i < e.NumWorker; i++ {
// Do some fake work
wg.Add(1)
go ad.GetResponses(params ,chan , &wg) //making HHTP call and returning the response in the channel
}
for v := range resChan {
//unifying all the response, and return the same as our response
switch v.Tyoe{
case A :
finalResponse.A = v
case B
finalResponse.B = v
}
}
return finalResponse
Request HTTP client
//i am using a global http client with custom transport , so that i can effectively use the resources
var client *http.Client
func init() {
tr := &http.Transport{
MaxIdleConnsPerHost: 1024,
TLSHandshakeTimeout: 0 * time.Second,
}
tr.MaxIdleConns = 100
tr.MaxConnsPerHost = 100
tr.MaxIdleConnsPerHost = 100
client = &http.Client{Transport: tr, Timeout: 10 * time.Second}
}
func GetResponses(params , chan ,wg){
res = client.Do(req)
chan <- res
}
So I have done some debugging and span monitoring , and turns out redis was the culprit in this. You can see this https://stackoverflow.com/a/70902382/9928176
To get an idea how I solved it

Golang Handlefunc (Runtime output display on browser) [duplicate]

I am trying to send a page response as soon as request is received, then process something, but I found the response does not get sent out "first" even though it is first in code sequence.In real life I have a page for uploading a excel sheet which gets saved into the database which takes time (50,0000+ rows) and would like to update to user progress. Here is a simplified example; (depending how much RAM you have you may need to add a couple zeros to counter to see result)
package main
import (
"fmt"
"net/http"
)
func writeAndCount(w http.ResponseWriter, r *http.Request) {
w.Write([]byte("Starting to count"))
for i := 0; i < 1000000; i++ {
if i%1000 == 0 {
fmt.Println(i)
}
}
w.Write([]byte("Finished counting"))
}
func main() {
http.HandleFunc("/", writeAndCount)
http.ListenAndServe(":8080", nil)
}
The original concept of the HTTP protocol is a simple request-response server-client computation model. There was no streaming or "continuous" client update support. It is (was) always the client who first contacted the server should it needed some kind of information.
Also since most web servers cache the response until it is fully ready (or a certain limit is reached–which is typically the buffer size), data you write (send) to the client won't be transmitted immediately.
Several techniques were "developed" to get around this "limitation" so that the server is able to notify the client about changes or progress, such as HTTP Long polling, HTTP Streaming, HTTP/2 Server Push or Websockets. You can read more about these in this answer: Is there a real server push over http?
So to achieve what you want, you have to step around the original "borders" of the HTTP protocol.
If you want to send data periodically, or stream data to the client, you have to tell this to the server. The easiest way is to check if the http.ResponseWriter handed to you implements the http.Flusher interface (using a type assertion), and if it does, calling its Flusher.Flush() method will send any buffered data to the client.
Using http.Flusher is only half of the solution. Since this is a non-standard usage of the HTTP protocol, usually client support is also needed to handle this properly.
First, you have to let the client know about the "streaming" nature of the response, by setting the ContentType=text/event-stream response header.
Next, to avoid clients caching the response, be sure to also set Cache-Control=no-cache.
And last, to let the client know that you might not send the response as a single unit (but rather as periodic updates or as a stream) and so that the client should keep the connection alive and wait for further data, set the Connection=keep-alive response header.
Once the response headers are set as the above, you may start your long work, and whenever you want to update the client about the progress, write some data and call Flusher.Flush().
Let's see a simple example that does everything "right":
func longHandler(w http.ResponseWriter, r *http.Request) {
flusher, ok := w.(http.Flusher)
if !ok {
http.Error(w, "Server does not support Flusher!",
http.StatusInternalServerError)
return
}
w.Header().Set("Content-Type", "text/event-stream")
w.Header().Set("Cache-Control", "no-cache")
w.Header().Set("Connection", "keep-alive")
start := time.Now()
for rows, max := 0, 50*1000; rows < max; {
time.Sleep(time.Second) // Simulating work...
rows += 10 * 1000
fmt.Fprintf(w, "Rows done: %d (%d%%), elapsed: %v\n",
rows, rows*100/max, time.Since(start).Truncate(time.Millisecond))
flusher.Flush()
}
}
func main() {
http.HandleFunc("/long", longHandler)
panic(http.ListenAndServe("localhost:8080", nil))
}
Now if you open http://localhost:8080/long in your browser, you will see an output "growing" by every second:
Rows done: 10000 (20%), elapsed: 1s
Rows done: 20000 (40%), elapsed: 2s
Rows done: 30000 (60%), elapsed: 3s
Rows done: 40000 (80%), elapsed: 4.001s
Rows done: 50000 (100%), elapsed: 5.001s
Also note that when using SSE, you should "pack" updates into SSE frames, that is you should start them with "data:" prefix, and end each frame with 2 newline chars: "\n\n".
"Literature" and further reading / tutorials
Read more about Server-sent events on Wikipedia.
See a Golang HTML5 SSE example.
See Golang SSE server example with client codes using it.
See w3school.com's turorial on Server-Sent Events - One Way Messaging.
You can check if the ResponseWriter is a http.Flusher, and if so, force the flush to network:
if f, ok := w.(http.Flusher); ok {
f.Flush()
}
However, bear in mind that this is a very unconventional HTTP handler. Streaming out progress messages to the response as if it were a terminal presents a few problems, particularly if the client is a web browser.
You might want to consider something more fitting with the nature of HTTP, such as returning a 202 Accepted response immediately, with a unique identifier the client can use to check on the status of processing using subsequent calls to your API.

Rate limit with golang.org/x/time/rate api request

I already created a function for limiting to 50 requests for API logins in one day.
var limit = 50
package middleware
import (
"log"
"net"
"net/http"
"sync"
"time"
"golang.org/x/time/rate"
)
// Create a custom request struct which holds the rate limiter for each
// visitor and the last time that the request was seen.
type request struct {
limiter *rate.Limiter
lastSeen time.Time
}
// Change the the map to hold values of the type request.
// defaultTime using 3 minutes
var requests = make(map[string]*request)
var mu sync.Mutex
func getRequest(ip string, limit int) *rate.Limiter {
mu.Lock()
defer mu.Unlock()
v, exists := requests[ip]
if !exists {
limiter := rate.NewLimiter(1, limit)
requests[ip] = &request{limiter, time.Now()}
return limiter
}
// Update the last seen time for the visitor.
v.lastSeen = time.Now()
return v.limiter
}
func throttle(next http.Handler, limit int) http.Handler {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
ip, _, err := net.SplitHostPort(r.RemoteAddr)
if err != nil {
log.Println(err.Error())
http.Error(w, "Internal Server Error", http.StatusInternalServerError)
return
}
limiter := getRequest(ip, limit)
fmt.Println(limiter.Allow())
if limiter.Allow() == false {
http.Error(w, http.StatusText(http.StatusTooManyRequests), http.StatusTooManyRequests)
return
}
next.ServeHTTP(w, r)
})
}
Is it correct?
Because when I try it, it still passes. The function limit is not working.
I doubt with NewLimiter()
limiter := rate.NewLimiter(1, limit)
Does it mean one user only can request login 50 requests per day? (I already read the docs, but I do not understand.)
From the rate docs:
func NewLimiter(r Limit, b int) *Limiter
NewLimiter returns a new Limiter that allows events up to rate r and
permits bursts of at most b tokens.
So the first parameter is the rate-limit, not the second. Burst is the number of requests you want to allow that occur faster than the rate-limit - typically one uses a value of 1 to disallow bursting, anything higher will let this number of requests in before the regular rate-limit kicks in. Anyway...
To create the rate.Limit for your needs, you can use the helper function rate.Every():
rt := rate.Every(24*time.Hour / 50)
limiter := rate.NewLimiter(rt, 1)
NewLimited(1, 50) means 1 request/second with a burst of up to 50 requests. It's a token bucket, which means that there are 50 tokens, each accepted API call uses up one token, and the tokens are regenerated at the given rate, up to burst. Your code is creating a limiter per IP address, so that's a limit per IP address (which I guess you are approximating as one IP address is one user).
If you're running on a single persistent server, and the server and code never restarts, then you may be able to get something like 50 requests/day per user by specifying a rate of 50 / (3600*24) and a burst of 50. (Note: 3600*24 is the number of seconds in a day). But the rate limiting package you're using is not designed for such coarse rate-limiting (on the order of requests per day) -- it's designed to prevent server overload under heavy traffic in the short term (on the order of requests per second).
You probably want a rate-limiter that works with a database or similar (perhaps using a token bucket scheme, since that can be implemented efficiently). Probably there's a package somewhere for that, but I don't know of one of the top of my head.

Measure upload speed when using http.ResponseBody

Is there a way to measure a client's download speed when uploading a large quantity of data using an http.ResponseWriter?
Update for context: I'm writing a streaming download endpoint for blob storage which stores blobs in chunks. The files are very large, so loading and buffering whole blobs is not feasible. Being able to monitor the buffer state, bytes written or similar would allow better scheduling of the chunk downloads.
E.g. when Write()ing to the response, is there a way to check how much data is already queued?
An example of the context, but not using a file object.
func downloadHandler(w http.ResponseWriter, req *http.Request, ps httprouter.Params) {
// Open some file.
f := os.Open("somefile.txt")
// Adjust the iteration speed of this loop to the client's download speed.
for
{
data := make([]byte, 1000)
count, err := f.Read(data)
if err != nil {
log.Fatal(err)
}
if count == 0 {
break
}
// Upload data chunk to client.
w.Write(data[:count])
}
}
You could implement a custom http.ResponseWriter that measures bytes sent, and calculates throughput.
There are likely packages to do similar things already. Google found this one (which I haven't used).

Why is my webserver in golang not handling concurrent requests?

This simple HTTP server contains a call to time.Sleep() that makes
each request take five seconds. When I try quickly loading multiple
tabs in a browser, it is obvious that each request
is queued and handled sequentially. How can I make it handle concurrent requests?
package main
import (
"fmt"
"net/http"
"time"
)
func serve(w http.ResponseWriter, r *http.Request) {
fmt.Fprintln(w, "Hello, world.")
time.Sleep(5 * time.Second)
}
func main() {
http.HandleFunc("/", serve)
http.ListenAndServe(":1234", nil)
}
Actually, I just found the answer to this after writing the question, and it is very subtle. I am posting it anyway, because I couldn't find the answer on Google. Can you see what I am doing wrong?
Your program already handles the requests concurrently. You can test it with ab, a benchmark tool which is shipped with Apache 2:
ab -c 500 -n 500 http://localhost:1234/
On my system, the benchmark takes a total of 5043ms to serve all 500 concurrent requests. It's just your browser which limits the number of connections per website.
Benchmarking Go programs isn't that easy by the way, because you need to make sure that your benchmark tool isn't the bottleneck and that it is also able to handle that many concurrent connections. Therefore, it's a good idea to use a couple of dedicated computers to generate load.
From Server.go , the go routine is spawned in the Serve function when a connection is accepted. Below is the snippet, :-
// Serve accepts incoming connections on the Listener l, creating a
// new service goroutine for each. The service goroutines read requests and
// then call srv.Handler to reply to them.
func (srv *Server) Serve(l net.Listener) error {
for {
rw, e := l.Accept()
if e != nil {
......
c, err := srv.newConn(rw)
if err != nil {
continue
}
c.setState(c.rwc, StateNew) // before Serve can return
go c.serve()
}
}
If you use xhr request, make sure that xhr instance is a local variable.
For example, xhr = new XMLHttpRequest() is a global variable. When you do parallel request with the same xhr variable you receive only one result. So, you must declare xhr locally like this var xhr = new XMLHttpRequest().

Resources