Building WebSocket Applications with Knative using Golang ¶
Published on: 2024-09-10
Building WebSocket Applications with Knative using Golang¶
Author: Matthias Weßendorf, Senior Principal Software Engineer @ Red Hat
Knative Serving supports multiple protocols for running serverless workloads, such as HTTP
, HTTP/2
, gRPC
or WebSocket
. The WebSocket protocol allows a full-duplex communication over one single TCP connection between client applications and servers. The WebSocket protocol starts as an HTTP "Upgrade" request and then switchs the connection from HTTP over to the WebSocket protocol, allowing real-time and bidirectional communication between the peers. In this article we show how to build a scalable WebSocket server with Knative.
Note
The sample application discussed here can be found in the Knative docs repository.
HTTP Upgrade¶
Since every WebSocket application starts with an HTTP Upgrade request, we need to define a Golang-based Webserver that handles classic HTTP requests. For the WebSocket implementation the gorilla/websocket
package is used, which offers a fast, well-tested and widely used WebSocket implementation for Golang. Below is HTTP part of our sample application:
package main
import (
"fmt"
"log"
"net/http"
"github.com/gorilla/websocket"
"github.com/knative/docs/code-samples/serving/websockets-go/pkg/handlers"
)
var upgrader = websocket.Upgrader{
ReadBufferSize: 1024,
WriteBufferSize: 1024,
CheckOrigin: func(r *http.Request) bool { return true },
}
func handleWebSocket(w http.ResponseWriter, r *http.Request) {
conn, err := upgrader.Upgrade(w, r, nil)
if err != nil {
log.Printf("Error upgrading to websocket: %v", err)
return
}
handlers.OnOpen(conn)
go func() {
defer handlers.OnClose(conn)
for {
messageType, message, err := conn.ReadMessage()
if err != nil {
handlers.OnError(conn, err)
break
}
handlers.OnMessage(conn, messageType, message)
}
}()
}
func main() {
http.HandleFunc("/ws", handleWebSocket)
fmt.Println("Starting server on :8080...")
if err := http.ListenAndServe(":8080", nil); err != nil {
log.Fatalf("Server error: %v", err)
}
}
The main
function starts the web server on the /ws
context and registers a handleWebSocket
handle function:
func main() {
http.HandleFunc("/ws", handleWebSocket)
fmt.Println("Starting server on :8080...")
if err := http.ListenAndServe(":8080", nil); err != nil {
log.Fatalf("Server error: %v", err)
}
}
The handleWebSocket
performs the protocol upgrade and assigns various websocket handler functions, such as OnOpen
or OnMessage
:
func handleWebSocket(w http.ResponseWriter, r *http.Request) {
conn, err := upgrader.Upgrade(w, r, nil)
if err != nil {
log.Printf("Error upgrading to websocket: %v", err)
return
}
handlers.OnOpen(conn)
go func() {
defer handlers.OnClose(conn)
for {
messageType, message, err := conn.ReadMessage()
if err != nil {
handlers.OnError(conn, err)
break
}
handlers.OnMessage(conn, messageType, message)
}
}()
}
Note
For a cleaner separation of concerns the different On...
handlers are located in separate handlers
package.
WebSocket handlers¶
The WebSocket application logic is located in the pkg/handlers/handlers.go
file and contains callbacks for each WebSocket event:
func OnOpen(conn *websocket.Conn) {
log.Printf("WebSocket connection opened: %v", conn.RemoteAddr())
}
func OnMessage(conn *websocket.Conn, messageType int, message []byte) {
log.Printf("Received message from %v: %s", conn.RemoteAddr(), string(message))
if err := conn.WriteMessage(messageType, message); err != nil {
log.Printf("Error sending message: %v", err)
}
}
func OnClose(conn *websocket.Conn) {
log.Printf("WebSocket connection closed: %v", conn.RemoteAddr())
conn.Close()
}
func OnError(conn *websocket.Conn, err error) {
log.Printf("WebSocket error from %v: %v", conn.RemoteAddr(), err)
}
In each of the handlers we log the remote address of the connected peer and of any incomming message through the OnMessage
handler we echo it back to the sender.
Testing the WebSocket server¶
Note
The easiest way to build and deploy Golang applications is using ko, as described on the actual sample application.
Once the Knative Serving application was deployed the the Kubernetes cluster, you can test it. For that you need to receive the actual URL for the service of the WebSocket application:
kubectl get ksvc
NAME URL LATESTCREATED LATESTREADY READY REASON
websocket-server http://websocket-server.default.svc.cluster.local websocket-server-00001 websocket-server-00001 True
Now we need a client to connect against our service. For that we launch a Linux container inside Kubernetes and run the wscat to connect against our WebSocket server application:
kubectl run --rm -i --tty wscat --image=monotykamary/wscat --restart=Never -- -c ws://websocket-server.default.svc.cluster.local/ws
Afterward you can chat with the WebSocket server like:
```If you don't see a command prompt, try pressing enter.
```connected (press CTRL+C to quit)
```> Hello
```< Hello
```>
The above is scaling to exactly one pod, since only one client was connected. Since Knative Serving allows you a dynamic scalling, a certain number of concurrent connections lead to a number of pods.
Below is an example of a scaled out application when running a larger number of concurrent requests against it:
k get pods
NAME READY STATUS RESTARTS AGE
websocket-server-00001-deployment-f785cbd65-42mrn 2/2 Running 0 16s
websocket-server-00001-deployment-f785cbd65-76bjr 2/2 Running 0 14s
websocket-server-00001-deployment-f785cbd65-98cwb 2/2 Running 0 18s
websocket-server-00001-deployment-f785cbd65-9fdbl 2/2 Running 0 18s
websocket-server-00001-deployment-f785cbd65-bpvjj 2/2 Running 0 20s
websocket-server-00001-deployment-f785cbd65-c6f47 2/2 Running 0 20s
websocket-server-00001-deployment-f785cbd65-fl6kk 2/2 Running 0 20s
websocket-server-00001-deployment-f785cbd65-gnffw 2/2 Running 0 20s
websocket-server-00001-deployment-f785cbd65-hqpfx 2/2 Running 0 20s
websocket-server-00001-deployment-f785cbd65-j4v9d 2/2 Running 0 18s
websocket-server-00001-deployment-f785cbd65-j72vk 2/2 Running 0 18s
websocket-server-00001-deployment-f785cbd65-k856w 2/2 Running 0 20s
websocket-server-00001-deployment-f785cbd65-kmmng 2/2 Running 0 20s
websocket-server-00001-deployment-f785cbd65-l4f2v 2/2 Running 0 20s
websocket-server-00001-deployment-f785cbd65-lpfr9 2/2 Running 0 14s
websocket-server-00001-deployment-f785cbd65-mn26w 2/2 Running 0 16s
websocket-server-00001-deployment-f785cbd65-mr98h 2/2 Running 0 18s
websocket-server-00001-deployment-f785cbd65-prjmj 2/2 Running 0 20s
websocket-server-00001-deployment-f785cbd65-v696v 2/2 Running 0 18s
websocket-server-00001-deployment-f785cbd65-ws5k9 2/2 Running 0 20s
websocket-server-00001-deployment-f785cbd65-xmszx 2/2 Running 0 18s
websocket-server-00001-deployment-f785cbd65-znhrr 2/2 Running 0 20s
Note
Depending on the target annotation you have (autoscaling.knative.dev/target
) you can scale based on num of connections.