在这个例子中，我们将演示如何优雅地关闭一个 Golang 应用程序。这个例子现在已经被很多人熟知。然而，不为人知的是，这个解决方案在Kubernetes环境中是否像在本地环境中一样有效。这完全取决于Kubernetes pod/部署配置中的terminationGracePeriodSeconds 选项。简而言之，确保上下文超时值小于terminationGracePeriodSeconds 选项的值。默认情况下，如果你不配置它，它被设置为30 秒，所以你的上下文超时值应该是n < 30 。
我们的应用程序允许长时间运行的请求/连接在40秒内完成。然后我们将terminationGracePeriodSeconds 设置为60。这将防止我们的应用程序被Kubernetes过早地杀死。通常情况下，不配置这个值，将上下文超时设置为10秒这样的值也是可以的。你会看到下面的日志：
结构
├── Makefile
├── deploy
│ └── k8s
│ ├── deployment.yaml
│ └── service.yaml
├── docker
│ └── dev
│ └── Dockerfile
└── main.go
文件
Makefile
.PHONY: run
run:
go run -race main.go
.PHONY: docker-push
docker-push:
docker build -t you/api:latest -f ./docker/dev/Dockerfile .
docker push you/api:latest
docker rmi you/api:latest
docker system prune --volumes --force
.PHONY: k8s-deploy
k8s-deploy:
kubectl apply -f deploy/k8s/deployment.yaml
kubectl apply -f deploy/k8s/service.yaml
Docker文件
FROM golang:1.15-alpine3.12 as build
WORKDIR /source
COPY . .
RUN CGO_ENABLED=0 go build -ldflags "-s -w" -o bin/main main.go
FROM alpine:3.12 as run
COPY --from=build /source/bin/main /main
ENTRYPOINT ["./main"]
main.go
package main
import (
"context"
"log"
"net/http"
"os"
"os/signal"
"syscall"
"time"
)
func main() {
// Create HTTP router.
rtr := http.NewServeMux()
rtr.HandleFunc("/", home)
// Create HTTP sever.
srv := &http.Server{
Addr: ":8080",
Handler: rtr,
}
// Start HTTP server.
go func() {
if err := srv.ListenAndServe(); err != nil {
log.Println("http server: listen and serve:", err)
}
}()
log.Println("app: started")
// Listen on application shutdown signals.
listener := make(chan os.Signal, 1)
signal.Notify(listener, os.Interrupt, syscall.SIGTERM)
log.Println("app: received a shutdown signal:", <-listener)
// Allow live connections a set period of time to complete.
ctx, cancel := context.WithTimeout(context.Background(), time.Second*40)
defer cancel()
// Shutdown HTTP server.
if err := srv.Shutdown(ctx); err != nil && err != http.ErrServerClosed {
log.Println("app: dirty shutdown:", err)
return
}
log.Println("app: clean shutdown")
}
// A dummy endpoint that simulates a long running request/connection/process.
func home(w http.ResponseWriter, _ *http.Request) {
log.Println("sleeping!")
time.Sleep(time.Second * 50)
log.Println("woke up!")
_, _ = w.Write([]byte("welcome home"))
}
service.yaml
apiVersion: v1
kind: Service
metadata:
name: api-service
spec:
type: NodePort
selector:
app: api
ports:
- protocol: TCP
port: 80
targetPort: 8080
部署.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
name: api-deployment
labels:
app: api
spec:
replicas: 1
selector:
matchLabels:
app: api
template:
metadata:
labels:
app: api
spec:
containers:
- name: golang
image: you/api:latest
ports:
- containerPort: 8080
terminationGracePeriodSeconds: 60
测试
当没有实时连接/请求或有但提前完成时，重新启动部署后的pod日志将如下所示。应用程序已被很好地关闭，所有请求都已被送达：
$ kubectl logs -f pod/api-deployment-66cb684477-7tjrf
2021/02/14 20:33:59 app: started
2021/02/14 20:34:17 sleeping!
2021/02/14 20:34:34 app: received a shutdown signal: terminated
2021/02/14 20:34:34 http server: listen and serve: http: Server closed
2021/02/14 20:35:07 woke up!
2021/02/14 20:35:08 app: clean shutdown
当有一个长期运行的连接/请求时，重新启动部署后，pod日志将看起来像下面这样。应用程序已被中断，长期运行的请求没有被提供：
$ kubectl logs -f pod/api-deployment-6f6d994c98-vv8jt
2021/02/14 20:26:13 app: started
2021/02/14 20:28:01 sleeping!
2021/02/14 20:28:10 app: received a shutdown signal: terminated
2021/02/14 20:28:10 http server: listen and serve: http: Server closed
2021/02/14 20:28:50 app: dirty shutdown: context deadline exceeded
当有一个长期运行的连接/请求，并且terminationGracePeriodSeconds 值小于40秒，这是上下文超时，重新启动部署后，pod日志将看起来像下面这样。这个pod已经被杀死了，所以甚至没有 "关闭 "日志!
$ kubectl logs -f pod/api-deployment-b8d5dc94c-wqb6s
2021/02/14 20:14:55 app: started
2021/02/14 20:16:26 sleeping!
2021/02/14 20:16:32 app: received a shutdown signal: terminated
2021/02/14 20:16:32 http server: listen and serve: http: Server closed
