Golang REST API 在 AWS EKS 上出现 CrashLoopBackOff 导致部署失败

编程并不是一个机械性的工作，而是需要有思考，有创新的工作，语法是固定的，但解决问题的思路则是依靠人的思维，这就需要我们坚持学习和更新自己的知识。今天golang学习网就整理分享《Golang REST API 在 AWS EKS 上出现 CrashLoopBackOff 导致部署失败》，文章讲解的知识点主要包括，如果你对Golang方面的知识点感兴趣，就不要错过golang学习网，在这可以对大家的知识积累有所帮助，助力开发能力的提升。

我正在尝试将一个用 golang 编写的简单 rest api 部署到 aws eks。

我使用 terraform 在 aws 上创建了一个 eks 集群，并向其应用了 aws 负载均衡器控制器 helm 图表。

集群中的所有资源如下所示：

namespace     name                                                ready   status    restarts   age
kube-system   pod/aws-load-balancer-controller-5947f7c854-fgwk2   1/1     running   0          75m
kube-system   pod/aws-load-balancer-controller-5947f7c854-gkttb   1/1     running   0          75m
kube-system   pod/aws-node-dfc7r                                  1/1     running   0          120m
kube-system   pod/aws-node-hpn4z                                  1/1     running   0          120m
kube-system   pod/aws-node-s6mng                                  1/1     running   0          120m
kube-system   pod/coredns-66cb55d4f4-5l7vm                        1/1     running   0          127m
kube-system   pod/coredns-66cb55d4f4-frk6p                        1/1     running   0          127m
kube-system   pod/kube-proxy-6ndf5                                1/1     running   0          120m
kube-system   pod/kube-proxy-s95qk                                1/1     running   0          120m
kube-system   pod/kube-proxy-vdrdd                                1/1     running   0          120m

namespace     name                                        type        cluster-ip      external-ip   port(s)         age
default       service/kubernetes                          clusterip   10.100.0.1              443/tcp         127m
kube-system   service/aws-load-balancer-webhook-service   clusterip   10.100.202.90           443/tcp         75m
kube-system   service/kube-dns                            clusterip   10.100.0.10             53/udp,53/tcp   127m

namespace     name                        desired   current   ready   up-to-date   available   node selector   age
kube-system   daemonset.apps/aws-node     3         3         3       3            3                     127m
kube-system   daemonset.apps/kube-proxy   3         3         3       3            3                     127m

namespace     name                                           ready   up-to-date   available   age
kube-system   deployment.apps/aws-load-balancer-controller   2/2     2            2           75m
kube-system   deployment.apps/coredns                        2/2     2            2           127m

namespace     name                                                      desired   current   ready   age
kube-system   replicaset.apps/aws-load-balancer-controller-5947f7c854   2         2         2       75m
kube-system   replicaset.apps/coredns-66cb55d4f4                        2         2         2       127m

我可以使用 go 和 docker 在本地运行应用程序。但在 aws eks 上发布它总是会抛出 crashloopbackoff。

运行 kubectl 描述 pod podname 显示：

name:         go-api-55d74b9546-dkk9g
namespace:    default
priority:     0
node:         ip-172-16-1-191.ec2.internal/172.16.1.191
start time:   tue, 15 mar 2022 07:04:08 -0700
labels:       app=go-api
              pod-template-hash=55d74b9546
annotations:  kubernetes.io/psp: eks.privileged
status:       running
ip:           172.16.1.195
ips:
  ip:           172.16.1.195
controlled by:  replicaset/go-api-55d74b9546
containers:
  go-api:
    container id:   docker://a4bc07b60c85fd308157d967d2d0d688d8eeccfe4c829102eb929ca82fb25595
    image:          saurabhmish/golang-hello:latest
    image id:       docker-pullable://saurabhmish/golang-hello@sha256:f79a495ad17710b569136f611ae3c8191173400e2cbb9cfe416e75e2af6f7874
    port:           3000/tcp
    host port:      0/tcp
    state:          waiting
      reason:       crashloopbackoff
    last state:     terminated
      reason:       error
      exit code:    1
      started:      tue, 15 mar 2022 07:09:50 -0700
      finished:     tue, 15 mar 2022 07:09:50 -0700
    ready:          false
    restart count:  6
    environment:    
    mounts:
      /var/run/secrets/kubernetes.io/serviceaccount from kube-api-access-jt4gp (ro)
conditions:
  type              status
  initialized       true 
  ready             false 
  containersready   false 
  podscheduled      true 
volumes:
  kube-api-access-jt4gp:
    type:                    projected (a volume that contains injected data from multiple sources)
    tokenexpirationseconds:  3607
    configmapname:           kube-root-ca.crt
    configmapoptional:       
    downwardapi:             true
qos class:                   besteffort
node-selectors:              
tolerations:                 node.kubernetes.io/not-ready:noexecute op=exists for 300s
                             node.kubernetes.io/unreachable:noexecute op=exists for 300s
events:
  type     reason     age                     from               message
  ----     ------     ----                    ----               -------
  normal   scheduled  7m31s                   default-scheduler  successfully assigned default/go-api-55d74b9546-dkk9g to ip-172-16-1-191.ec2.internal
  normal   pulled     7m17s                   kubelet            successfully pulled image "saurabhmish/golang-hello:latest" in 12.77458991s
  normal   pulled     7m16s                   kubelet            successfully pulled image "saurabhmish/golang-hello:latest" in 110.127771ms
  normal   pulled     7m3s                    kubelet            successfully pulled image "saurabhmish/golang-hello:latest" in 109.617419ms
  normal   created    6m37s (x4 over 7m17s)   kubelet            created container go-api
  normal   started    6m37s (x4 over 7m17s)   kubelet            started container go-api
  normal   pulled     6m37s                   kubelet            successfully pulled image "saurabhmish/golang-hello:latest" in 218.952336ms
  normal   pulling    5m56s (x5 over 7m30s)   kubelet            pulling image "saurabhmish/golang-hello:latest"
  normal   pulled     5m56s                   kubelet            successfully pulled image "saurabhmish/golang-hello:latest" in 108.105083ms
  warning  backoff    2m28s (x24 over 7m15s)  kubelet            back-off restarting failed container

运行 kubectl 日志 podname 和 kubectl 日志 podname -c go-api 显示 standard_init_linux.go:228: exec 用户进程导致：exec 格式错误

清单：

go-deploy.yaml（这是带有文档的 docker hub 映像）

---
apiversion: apps/v1
kind: deployment
metadata:
  name: go-api
  labels:
    app: go-api
spec:
  replicas: 2
  selector:
    matchlabels:
      app: go-api
  strategy: {}
  template:
    metadata:
      labels:
        app: go-api
    spec:
      containers:
      - name: go-api
        image: saurabhmish/golang-hello:latest
        ports:
          - containerport: 3000
        resources: {}

go-service.yaml

---
kind: Service
apiVersion: v1
metadata:
  name: go-api
spec:
  selector:
    app: go-api
  type: NodePort
  ports:
  - protocol: TCP
    port: 80
    targetPort: 3000

如何修复此错误？

正确答案

将其发布为社区 wiki 以获得更好的可见性。 请随意扩展它。

感谢@david maze，他指出了解决方案。有一篇文章“从 mac m1 (arm) 构建兼容 intel64 的 docker 映像”（作者：beppe catanese）here。
本文很好地描述了根本问题。

您正在 arm 架构 (mac m1) 上进行开发/构建，但您将 docker 映像部署到基于 x86-64 架构的 kubernetes 集群。

解决方案：

选项 a：使用 buildx

Buildx 是一个 docker 插件，除其他功能外，它还允许为各种目标平台构建映像。

$ docker buildx build --platform linux/amd64 -t myapp .

选项 b：设置 docker_default_platform

docker_default_platform 环境变量允许为采用 --platform 标志的命令设置默认平台。

export DOCKER_DEFAULT_PLATFORM=linux/amd64

crashloopbackoff 意味着您有一个 pod 启动、崩溃、再次启动，然后再次崩溃。

也许错误来自应用程序本身，它无法连接到数据库、redis...

您可能会在这里找到有用的东西：

My kubernetes pods keep crashing with "CrashLoopBackOff" but I can't find any log

好了，本文到此结束，带大家了解了《Golang REST API 在 AWS EKS 上出现 CrashLoopBackOff 导致部署失败》，希望本文对你有所帮助！关注golang学习网公众号，给大家分享更多Golang知识！