乔克视界

主题

2025-07-222025-07-23博客运维
2337
12.5m
0

NGINX容器热更新 ​

Nginx 作为 WEB 服务器被广泛使用。其自身支持热更新，在修改配置文件后，使用nginx -s reload命令可以不停服务重新加载配置。然而对于 Dockerize 的 Nginx 来说，如果每次都进到容器里执行对应命令去实现配置重载，这个过程是很痛苦的。本文介绍了一种 kubernetes 集群下 nginx 的热更新方案。

首先我们创建正常的一个 nginx 资源，资源清单如下：

yaml
apiVersion: v1
kind: ConfigMap
metadata:
  name: nginx-config
data:
  default.conf: |-
    server {
      server_name localhost;
      listen 80 default_server;

      location = /healthz {
        add_header Content-Type text/plain;
        return 200 'ok';
      }

      location / {
          root   /usr/share/nginx/html;
          index  index.html index.htm;
      }

      error_page   500 502 503 504  /50x.html;

      location = /50x.html {
          root   /usr/share/nginx/html;
      }
    }
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: my-app
spec:
  replicas: 1
  selector:
    matchLabels:
      app: my-app
  template:
    metadata:
      labels:
        app: my-app
    spec:
      containers:
        - name: my-app
          image: nginx
          imagePullPolicy: IfNotPresent
          volumeMounts:
            - name: nginx-config
              mountPath: /etc/nginx/conf.d
      volumes:
        - name: nginx-config
          configMap:
            name: nginx-config

然后创建资源对象。

yaml
# kubectl get pod -o wide
NAME                     READY   STATUS    RESTARTS   AGE    IP               NODE         NOMINATED NODE   READINESS GATES
my-app-9bdd6cbbc-x9gnt   1/1     Running   0          112s   192.168.58.197   k8s-node02   <none>           <none>

然后我们访问 pod 资源，如下：

yaml
# curl -I 192.168.58.197
HTTP/1.1 200 OK
Server: nginx/1.17.10
Date: Tue, 26 May 2020 06:18:18 GMT
Content-Type: text/html
Content-Length: 612
Last-Modified: Tue, 14 Apr 2020 14:19:26 GMT
Connection: keep-alive
ETag: "5e95c66e-264"
Accept-Ranges: bytes

现在我们来更新一下 ConfigMap，也就是更改配置文件如下：

yaml
apiVersion: v1
kind: ConfigMap
metadata:
  name: nginx-config
data:
  default.conf: |-
    server {
      server_name localhost;
      listen 8080 default_server;

      location = /healthz {
        add_header Content-Type text/plain;
        return 200 'ok';
      }

      location / {
          root   /usr/share/nginx/html;
          index  index.html index.htm;
      }

      error_page   500 502 503 504  /50x.html;

      location = /50x.html {
          root   /usr/share/nginx/html;
      }
    }

等待数秒...

然后我们可以看到 nginx pod 里的配置信息已经更改为如下：

bash
# kubectl exec -it my-app-9bdd6cbbc-x9gnt -- /bin/bash
root@my-app-9bdd6cbbc-x9gnt:/# cat  /etc/nginx/conf.d/default.conf
server {
server_name localhost;
listen 8080 default_server;

location = /healthz {
add_header Content-Type text/plain;
return 200 'ok';
}

location / {
root   /usr/share/nginx/html;
index  index.html index.htm;
}

error_page   500 502 503 504  /50x.html;

location = /50x.html {
root   /usr/share/nginx/html;
}
}
root@my-app-9bdd6cbbc-x9gnt:/#

这时候我们访问 8080 是不通的，访问 80 是没问题的，如下：

bash
[root@k8s-master nginx]# curl -I 192.168.58.197
HTTP/1.1 200 OK
Server: nginx/1.17.10
Date: Tue, 26 May 2020 06:21:05 GMT
Content-Type: text/html
Content-Length: 612
Last-Modified: Tue, 14 Apr 2020 14:19:26 GMT
Connection: keep-alive
ETag: "5e95c66e-264"
Accept-Ranges: bytes

[root@k8s-master nginx]# curl -I 192.168.58.197:8080
curl: (7) Failed connect to 192.168.58.197:8080; Connection refused

我们可以看到，我们需要的配置文件已经更新的，但是并没有使用上，pod 里的 nginx 也没有重载配置文件，这时候如果我们重新部署 Pod，资源对象肯定就生效了。

但是这并不我们想要的效果，我们希望配置文件更改了，服务也跟着 reload，并不需要我们手动的去干预。

目前有三种方法：

  • 应用本身可以检测配置文件，然后自动 reload
  • 给 Pod 增加一个 sidecar，用它来检测配置文件
  • 第三方组件 reloader，在 deployment 的 annotations 增加字段reloader.stakater.com/auto: "true",即可检测 configmap 的更改来重启 pod

应用本身检测的话这里就不做介绍了。这里主要来实验一下第 2，3 种方法

一、以 sidecar 形式 ​

1.1、方法 ​

  • Kubernetes 集群中部署 Nginx Pod。该 Pod 包含两个 Container，一个是 nginx container，实现 nginx 自身的功能；另一个是 nginx-reloader container，负责实时监测目标 configmap 的变化，当发现 configmap 更新以后，会主动向 nginx 的 master 进程发送 HUP 信号，实现配置的热加载。
  • 配置文件是通过 ConfigMap 的形式挂载到 Nginx Pod 上，两个 Container 共享该 ConfigMap。
  • 依赖 K8s 集群的 shareProcessNamespace 特性（版本需在 1.12 之后），两个 Container 需要在 Pod 中共享进程名字空间。

1.2、实现 ​

1.2.1、镜像制作 ​

（1）、主容器使用官方 NG 容器即可

（2）、sidecar 容器制作

Dockerfile 如下：

dockerfile
FROM golang:1.12.0 as build

RUN go get github.com/fsnotify/fsnotify
RUN go get github.com/shirou/gopsutil/process
RUN mkdir -p /go/src/app
ADD main.go /go/src/app/
WORKDIR /go/src/app
RUN CGO_ENABLED=0 GOOS=linux go build -a -o nginx-reloader .

# main image
FROM nginx:1.14.2-alpine

COPY --from=build /go/src/app/nginx-reloader /

CMD ["/nginx-reloader"]

main.go 脚本如下：

go
package main

import (
    "log"
    "os"
    "path/filepath"
    "syscall"

    "github.com/fsnotify/fsnotify"
    proc "github.com/shirou/gopsutil/process"
)

const (
    nginxProcessName = "nginx"
    defaultNginxConfPath = "/etc/nginx"
    watchPathEnvVarName = "WATCH_NGINX_CONF_PATH"
)

var stderrLogger = log.New(os.Stderr, "error: ", log.Lshortfile)
var stdoutLogger = log.New(os.Stdout, "", log.Lshortfile)

func getMasterNginxPid() (int, error) {
    processes, processesErr := proc.Processes()
    if processesErr != nil {
        return 0, processesErr
    }

    nginxProcesses := map[int32]int32{}

    for _, process := range processes {
        processName, processNameErr := process.Name()
        if processNameErr != nil {
            return 0, processNameErr
        }

        if processName == nginxProcessName {
            ppid, ppidErr := process.Ppid()

            if ppidErr != nil {
                return 0, ppidErr
            }

            nginxProcesses[process.Pid] = ppid
        }
    }

    var masterNginxPid int32

    for pid, ppid := range nginxProcesses {
        if ppid == 0 {
            masterNginxPid = pid

            break
        }
    }

    stdoutLogger.Println("found master nginx pid:", masterNginxPid)

    return int(masterNginxPid), nil
}

func signalNginxReload(pid int) error {
    stdoutLogger.Printf("signaling master nginx process (pid: %d) -> SIGHUP\n", pid)
    nginxProcess, nginxProcessErr := os.FindProcess(pid)

    if nginxProcessErr != nil {
        return nginxProcessErr
    }

    return nginxProcess.Signal(syscall.SIGHUP)
}

func main() {
    watcher, watcherErr := fsnotify.NewWatcher()
    if watcherErr != nil {
        stderrLogger.Fatal(watcherErr)
    }
    defer watcher.Close()

    done := make(chan bool)
    go func() {
        for {
            select {
            case event, ok := <-watcher.Events:
                if !ok {
                    return
                }

                if event.Op&fsnotify.Create == fsnotify.Create {
                    if filepath.Base(event.Name) == "..data" {
                        stdoutLogger.Println("config map updated")

                        nginxPid, nginxPidErr := getMasterNginxPid()
                        if nginxPidErr != nil {
                            stderrLogger.Printf("getting master nginx pid failed: %s", nginxPidErr.Error())

                            continue
                        }

                        if err := signalNginxReload(nginxPid); err != nil {
                            stderrLogger.Printf("signaling master nginx process failed: %s", err)
                        }
                    }
                }
            case err, ok := <-watcher.Errors:
                if !ok {
                    return
                }
                stderrLogger.Printf("received watcher.Error: %s", err)
            }
        }
    }()

    pathToWatch, ok := os.LookupEnv(watchPathEnvVarName)
    if !ok {
        pathToWatch = defaultNginxConfPath
    }

    stdoutLogger.Printf("adding path: `%s` to watch\n", pathToWatch)

    if err := watcher.Add(pathToWatch); err != nil {
        stderrLogger.Fatal(err)
    }
    <-done
}

1.2.2、部署 NG ​

（1）、NG 的配置以 configMap 进行部署：

nginx-config.yaml

yaml
// nginx-config.yaml
apiVersion: v1
kind: ConfigMap
metadata:
  name: nginx-config
data:
  default.conf: |-
      server {
        server_name localhost;
        listen 80 default_server;

        location = /healthz {
          add_header Content-Type text/plain;
          return 200 'ok';
        }

        location / {
            root   /usr/share/nginx/html;
            index  index.html index.htm;
        }

        error_page   500 502 503 504  /50x.html;

        location = /50x.html {
            root   /usr/share/nginx/html;
        }
      }

（2）、NG 的 Deployment 清单(需打开共享进程命名空间特性：shareProcessNamespace: true)：

nginx-deploy.yaml

yaml
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx
spec:
  replicas: 1
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      name: nginx
      labels:
        app: nginx
    spec:
      shareProcessNamespace: true
      containers:
        - name: nginx
          image: nginx
          imagePullPolicy: IfNotPresent
          volumeMounts:
            - name: nginx-config
              mountPath: /etc/nginx/conf.d
              readOnly: true
        - name: nginx-reloader
          image: registry.cn-hangzhou.aliyuncs.com/rookieops/nginx-reloader:v1
          imagePullPolicy: IfNotPresent
          env:
            - name: WATCH_NGINX_CONF_PATH
              value: /etc/nginx/conf.d
          volumeMounts:
            - name: nginx-config
              mountPath: /etc/nginx/conf.d
              readOnly: true
      volumes:
        - name: nginx-config
          configMap:
            name: nginx-config

手动修改 configmap 后，reloader 监测到 configmap 变化，会主动向 nginx 主进程发起 HUP 信号，实现配置热更新。

二、第三方插件 reloader ​

项目地址：https://github.com/stakater/Reloader

资源清单如下，我修改了镜像地址：

yaml
---
# Source: reloader/templates/clusterrole.yaml

apiVersion: rbac.authorization.k8s.io/v1beta1
kind: ClusterRole
metadata:
  labels:
    app: reloader-reloader
    chart: "reloader-v0.0.58"
    release: "reloader"
    heritage: "Tiller"
  name: reloader-reloader-role
  namespace: default
rules:
  - apiGroups:
      - ""
    resources:
      - secrets
      - configmaps
    verbs:
      - list
      - get
      - watch
  - apiGroups:
      - "apps"
    resources:
      - deployments
      - daemonsets
      - statefulsets
    verbs:
      - list
      - get
      - update
      - patch
  - apiGroups:
      - "extensions"
    resources:
      - deployments
      - daemonsets
    verbs:
      - list
      - get
      - update
      - patch

---
# Source: reloader/templates/clusterrolebinding.yaml

apiVersion: rbac.authorization.k8s.io/v1beta1
kind: ClusterRoleBinding
metadata:
  labels:
    app: reloader-reloader
    chart: "reloader-v0.0.58"
    release: "reloader"
    heritage: "Tiller"
  name: reloader-reloader-role-binding
  namespace: default
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: reloader-reloader-role
subjects:
  - kind: ServiceAccount
    name: reloader-reloader
    namespace: default

---
# Source: reloader/templates/deployment.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  labels:
    app: reloader-reloader
    chart: "reloader-v0.0.58"
    release: "reloader"
    heritage: "Tiller"
    group: com.stakater.platform
    provider: stakater
    version: v0.0.58

  name: reloader-reloader
spec:
  replicas: 1
  revisionHistoryLimit: 2
  selector:
    matchLabels:
      app: reloader-reloader
      release: "reloader"
  template:
    metadata:
      labels:
        app: reloader-reloader
        chart: "reloader-v0.0.58"
        release: "reloader"
        heritage: "Tiller"
        group: com.stakater.platform
        provider: stakater
        version: v0.0.58

    spec:
      containers:
        - env:
          image: "registry.cn-hangzhou.aliyuncs.com/rookieops/stakater-reloader:v0.0.58"
          imagePullPolicy: IfNotPresent
          name: reloader-reloader
          args:
      serviceAccountName: reloader-reloader

---
# Source: reloader/templates/role.yaml
---
# Source: reloader/templates/rolebinding.yaml
---
# Source: reloader/templates/service.yaml
---
# Source: reloader/templates/serviceaccount.yaml

apiVersion: v1
kind: ServiceAccount
metadata:
  labels:
    app: reloader-reloader
    chart: "reloader-v0.0.58"
    release: "reloader"
    heritage: "Tiller"
  name: reloader-reloader

然后部署资源，结果如下：

yaml
kubectl get pod
NAME                               READY   STATUS    RESTARTS   AGE
my-app-9bdd6cbbc-x9gnt             1/1     Running   0          38m
reloader-reloader-ff767bb8-cpzgz   1/1     Running   0          56s

然后给 deployment 增加一个 annotations。如下：

yaml
kubectl patch deployments.apps my-app -p '{"metadata": {"annotations": {"reloader.stakater.com/auto": "true"}}}'

然后我们更改 configMap 清单，重新 apply 过后，我们可以看到 pod 会删除重启，如下：

yaml
kubectl get pod -o wide
NAME                               READY   STATUS        RESTARTS   AGE     IP               NODE         NOMINATED NODE   READINESS GATES
my-app-7c4fc77f5f-w4mbn            1/1     Running       0          3s      192.168.58.202   k8s-node02   <none>           <none>
my-app-df6fbdb67-bnftb             1/1     Terminating   0          35s     192.168.58.201   k8s-node02   <none>           <none>
reloader-reloader-ff767bb8-cpzgz   1/1     Running       0          3m47s   192.168.85.195   k8s-node01   <none>           <none>

然后我们 curl pod 也可以通了，如下：

yaml
# curl 192.168.58.202:8080 -I
HTTP/1.1 200 OK
Server: nginx/1.17.10
Date: Tue, 26 May 2020 06:58:38 GMT
Content-Type: text/html
Content-Length: 612
Last-Modified: Tue, 14 Apr 2020 14:19:26 GMT
Connection: keep-alive
ETag: "5e95c66e-264"
Accept-Ranges: bytes

三、附加 ​

附加一个 sidecar 形式的 python 脚本

python
#!/usr/bin/env python
# -*- encoding: utf8 -*-
"""
需求：nginx配置文件变化，自动更新配置文件，类似nginx -s reload
实现：
    1、用pyinotify实时监控nginx配置文件变化
    2、如果配置文件变化，给系统发送HUP来reload nginx
"""
import os
import re
import pyinotify
import logging
from threading import Timer

# Param
LOG_PATH = "/root/python/log"
CONF_PATHS = [
  "/etc/nginx",
]
DELAY = 5
SUDO = False
RELOAD_COMMAND = "nginx -s reload"
if SUDO:
  RELOAD_COMMAND = "sudo " + RELOAD_COMMAND

# Log
logger = logging.getLogger(__name__)
logger.setLevel(level = logging.INFO)
log_handler = logging.FileHandler(LOG_PATH)
log_handler.setLevel(logging.INFO)
log_formatter = logging.Formatter('%(asctime)s - %(levelname)s - %(message)s')
log_handler.setFormatter(log_formatter)
logger.addHandler(log_handler)

# Reloader
def reload_nginx():
  os.system(RELOAD_COMMAND)
  logger.info("nginx is reloaded")

t = Timer(DELAY, reload_nginx)

def trigger_reload_nginx(pathname, action):
  logger.info("nginx monitor is triggered because %s is %s" % (pathname, action))
  global t
  if t.is_alive():
    t.cancel()
    t = Timer(DELAY, reload_nginx)
    t.start()
  else:
    t = Timer(DELAY, reload_nginx)
    t.start()

events = pyinotify.IN_MODIFY | pyinotify.IN_CREATE | pyinotify.IN_DELETE

watcher = pyinotify.WatchManager()
watcher.add_watch(CONF_PATHS, events, rec=True, auto_add=True)

class EventHandler(pyinotify.ProcessEvent):
  def process_default(self, event):
    if event.name.endswith(".conf"):
      if event.mask == pyinotify.IN_CREATE:
        action = "created"
      if event.mask == pyinotify.IN_MODIFY:
        action = "modified"
      if event.mask == pyinotify.IN_DELETE:
        action = "deleted"
      trigger_reload_nginx(event.pathname, action)

handler = EventHandler()
notifier = pyinotify.Notifier(watcher, handler)

# Start
logger.info("Start Monitoring")
notifier.loop()
最近更新