实现GAE和GKE k8s cron job的秒级调度单位 #73

mrdulin · 2019-04-08T13:22:40Z

实现GAE和GKE k8s cron job的秒级调度单位

问题

如何实现Google App Engine和Google Kubernetes Engine定时作业（cron job）的秒级调度（schedule）单位？

不同于node-schedule拥有秒级调度单位，可以将定时作业精确在某一秒执行。GAE和GKE的定时作业只能精确到分钟（min）。

在GAE的定时作业文档中，schedule格式可以知道，时间单位的有效值包括：

minutes 或 mins
hours

在GKE的定时作业文档中，CronJob 使用所需的 schedule 字段，该字段接受 Unix 标准 crontab 格式的时间。所有 CronJob 时间均采用世界协调时间 (UTC)：

第一个值表示分钟（0 到 59 之间）
第二个值表示小时（0 到 23 之间）
第三个值表示一个月中的某天（1 到 31 之间）
第四个值表示月份（1 到 12 之间）
第五个值表示星期（0 到 6 之间）

以及k8s Cron Job文档中的调度(schedule)都是Unix标准crontab格式的时间，GKE使用kubernetes，简称k8s，来管理compute engine集群（kubernetes集群）。

解决方案

`GAE` cron job

我们可以先尝试设置GAE定时作业的schedule为every 1 seconds，看看是什么效果。

server.js:

const express = require('express');
const pkg = require('./package.json');
const app = express();

function validateCronRequest(req, res, next) {
  console.log('X-Appengine-Cron', req.get('X-Appengine-Cron'), typeof req.get('X-Appengine-Cron'));
  if (req.get('X-Appengine-Cron') !== 'true') {
    return res.status(403);
  }
  next();
}

app.get('/version', (req, res) => {
  res.send(`version: ${pkg.version}`);
});

app.get('/cronjob/sync', validateCronRequest, (req, res) => {
  res.sendStatus(200);
});

const PORT = process.env.PORT || 8080;
app.listen(PORT, () => {
  console.log(`Server listening on port ${PORT}...`);
});

app.yaml：

runtime: nodejs8
service: cronjob-sec-unit

cron.yaml:

cron:
  - description: 'sync job'
    url: /cronjob/sync
    schedule: every 1 seconds
    target: cronjob-sec-unit

首先部署GAE service:

gcloud app deploy

部署成功后，去GCP console - App Engine查看

紧接着部署GAE定时作业:

gcloud app deploy ./cron.yaml

报错

ERROR: (gcloud.app.deploy) An error occurred while parsing file: [/Users/ldu020/workspace/nodejs-gcp/src/app-engine/cron-jobs-second-unit-schedule/cron.yaml]
Unable to assign value 'every 1 seconds' to attribute 'schedule':
schedule 'every 1 seconds' failed to parse: line 1:8 mismatched input u'second' expecting set None
in "/Users/ldu020/workspace/nodejs-gcp/src/app-engine/cron-jobs-second-unit-schedule/cron.yaml", line 4, column 15

错误说“不能将every 1 seconds赋值给schedule，解析every 1 seconds失败”。可见，GAE定时作业schedule并不支持秒级单位。

修改cron.yaml，将schedule设置为every 1 mins：

cron:
  - description: 'sync job'
    url: /cronjob/sync
    schedule: every 1 mins
    target: cronjob-sec-unit

再次部署，部署成功后查看GCP控制台:

查看该定时作业在Stackdriver Logging中的logs，验证该定时作业运行是否正常。

通过日志的时间戳可以看到，该定时作业每分钟执行一次。

查看App Engine部署的service的logs:

该service运行正常，路由/cronjob/sync每分钟被定时作业调用一次。

接下来实现秒级调度单位的定时作业，本质上是将定时作业的最小调度单位——1分钟，再次划分成一系列秒级时间序列。对Node.js异步流程控制库async熟悉的同学可以马上想到用timeSeries函数。我们这里将定时作业设置为10秒执行一次，修改后的代码如下：

server.js:

const express = require('express');
const async = require('async');

const pkg = require('./package.json');
const app = express();

const OUTER_SCHEDULE = 60; // 1 minute
const INNER_SCHEDULE = 10; // 10 seconds

function validateCronRequest(req, res, next) {
  // console.log('X-Appengine-Cron', req.get('X-Appengine-Cron'), typeof req.get('X-Appengine-Cron'));
  if (process.env.NODE_ENV === 'production') {
    if (req.get('X-Appengine-Cron') !== 'true') {
      return res.status(403);
    }
  }

  next();
}

app.get('/version', (req, res) => {
  res.send(`version: ${pkg.version}`);
});

app.get('/cronjob/sync', validateCronRequest, async (req, res) => {
  // Update database every 1 minute
  // await updateDB();

  // Update database every 10 seconds
  try {
    await updateDBTimeSeries();
  } catch (error) {
    console.error(error);
  }
  res.sendStatus(200);
});

function sleep(ms) {
  return new Promise((resolve) => setTimeout(resolve, ms));
}

async function updateDB() {
  return new Promise((resolve) => {
    process.nextTick(() => {
      console.log('Update database success.');
      resolve();
    });
  });
}

async function updateDBEveryTenSeconds(item, n, next) {
  // console.log('item: ', item);
  await updateDB();
  if (item < n - 1) {
    await sleep(INNER_SCHEDULE * 1000);
  }
  next();
}

async function updateDBTimeSeries() {
  const n = OUTER_SCHEDULE / INNER_SCHEDULE;
  return new Promise((resolve, reject) => {
    async.timesSeries(
      n,
      (item, next) => updateDBEveryTenSeconds(item, n, next),
      (error, results) => {
        if (error) {
          return reject(error);
        }
        console.log('A round update database operations done.');
        resolve(results);
      }
    );
  });
}

const PORT = process.env.PORT || 8080;
app.listen(PORT, () => {
  console.log(`Server listening on port ${PORT}...`);
});

部署后，查看App Engine service的日志：

可以看到，我们的调度任务每10秒执行一次，通过async.timeSeries方法将GAE定时作业的schedule（1分钟），划分出了6段。

GKE cron job

同理，我们也可以在GKE的定时作业上实现秒级调度单位。

src/index.js:

const async = require('async');

const OUTER_SCHEDULE = 60;
const INNER_SCHEDULE = 10;

async function main() {
  try {
    await updateDBTimeSeries();
    console.log('A round update database operations done.');
  } catch (error) {
    console.error(error);
  }
}

function sleep(ms) {
  return new Promise((resolve) => setTimeout(resolve, ms));
}

async function updateDB() {
  return new Promise((resolve) => {
    process.nextTick(() => {
      console.log('Update database success.');
      resolve();
    });
  });
}

async function updateDBEveryTenSeconds(item, n, next) {
  await updateDB();
  if (item < n - 1) {
    await sleep(INNER_SCHEDULE * 1000);
  }
  next();
}

async function updateDBTimeSeries() {
  const n = OUTER_SCHEDULE / INNER_SCHEDULE;
  return new Promise((resolve, reject) => {
    async.timesSeries(
      n,
      (item, next) => updateDBEveryTenSeconds(item, n, next),
      (error, results) => {
        if (error) {
          return reject(error);
        }
        resolve(results);
      }
    );
  });
}

main();

示例代码和GAE例子的代码差不多，去掉了express启动的HTTP server的代码。

k8s/cronjobs/update-db.yaml:

apiVersion: batch/v1beta1
kind: CronJob
metadata:
  name: update-db
spec:
  schedule: '*/1 * * * *'
  startingDeadlineSeconds: 10
  concurrencyPolicy: Forbid
  successfulJobsHistoryLimit: 1
  failedJobsHistoryLimit: 1
  jobTemplate:
    spec:
      template:
        metadata:
          labels:
            run: update-db
        spec:
          restartPolicy: Never
          containers:
            - name: cron-job-second-unit-schedule
              image: asia.gcr.io/shadowsocks-218808/cron-job-second-unit-schedule:1.1
              env:
                - name: NODE_ENV
                  value: 'production'
                - name: K8S_POD_NAME
                  valueFrom:
                    fieldRef:
                      fieldPath: metadata.name
              ports:
                - containerPort: 8080
                  protocol: TCP

设置schedule为*/1 * * * *，表示每分钟执行一次作业，具体执行什么作业？执行容器CMD指令指定的启动脚本，即npm start，即node ./src/index.js。

Dockerfile:

FROM node:8.9-alpine

WORKDIR /app
COPY ./ /app/

RUN apk update \
  && apk add curl --no-cache \
  && npm i -g npm@latest \
  && npm i --production

CMD [ "npm", "start" ]

package.json:

"scripts": {
    "start": "node ./src/index.js"
},

下面关于打包docker image，还有GKE的使用和k8s cron job的部署简单说下。

打包docker image:

docker build -t cron-job-second-unit-schedule:1.1 .

设置docker tag

docker tag cron-job-second-unit-schedule:1.1 asia.gcr.io/shadowsocks-218808/cron-job-second-unit-schedule:1.1

push该docker image到GCR registry:

docker push asia.gcr.io/shadowsocks-218808/cron-job-second-unit-schedule:1.1
The push refers to repository [asia.gcr.io/shadowsocks-218808/cron-job-second-unit-schedule]
2a9069409556: Pushed
4a18dd75bba0: Pushed
fbe45ea1c2a6: Layer already exists
f846841ed47f: Layer already exists
0198944a9875: Layer already exists
9dfa40a0da3b: Layer already exists
1.1: digest: sha256:81e4b2b5ddda93781bf368372d8697cad5619ca7ffa763cdd1eb0a49893622d1 size: 1577

使用gcloud命令行工具创建kubernetes cluster:

☁  cron-job-second-unit-schedule [master] ⚡  gcloud container clusters create nodejs-gcp --num-nodes=3

Creating cluster nodejs-gcp in us-west1-a... Cluster is being health-checked (master is healthy)...done.
Created [https://container.googleapis.com/v1/projects/shadowsocks-218808/zones/us-west1-a/clusters/nodejs-gcp].
To inspect the contents of your cluster, go to: https://console.cloud.google.com/kubernetes/workload_/gcloud/us-west1-a/nodejs-gcp?project=shadowsocks-218808
kubeconfig entry generated for nodejs-gcp.
NAME        LOCATION    MASTER_VERSION  MASTER_IP      MACHINE_TYPE   NODE_VERSION   NUM_NODES  STATUS
nodejs-gcp  us-west1-a  1.11.7-gke.12   35.199.144.38  n1-standard-1  1.11.7-gke.12  3          RUNNING

为kubectl命令设置默认集群，将kubectl的context设置为新创建的这个cluster

☁  cron-job-second-unit-schedule [master] ⚡  gcloud container clusters get-credentials nodejs-gcp
Fetching cluster endpoint and auth data.
kubeconfig entry generated for nodejs-gcp.

部署k8s cron job:

☁  cron-job-second-unit-schedule [master] ⚡  kaf ./k8s/cronjobs/update-db.yaml
cronjob.batch "update-db" created

部署完成后，去GCP控制台查看：

查看update-db定时作业的container logs:

可以看到该定时作业每10秒执行一次。

源码地址

GAE: https://github.com/mrdulin/nodejs-gcp/tree/master/src/app-engine/cron-jobs-second-unit-schedule
GKE: https://github.com/mrdulin/nodejs-gcp/tree/master/src/kubernetes-engine/cron-job-second-unit-schedule

顺便推荐个工具：https://crontab.guru/ - 输入*/1 * * * *这样的crontab风格的schedule，会输出文字表述的时间。

The text was updated successfully, but these errors were encountered:

mrdulin added GCP google-cloud-platform Node.js 编程软件工具技术文档 and removed 软件工具 labels Apr 8, 2019

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

实现GAE和GKE k8s cron job的秒级调度单位 #73

实现GAE和GKE k8s cron job的秒级调度单位 #73

mrdulin commented Apr 8, 2019 •

edited

实现GAE和GKE k8s cron job的秒级调度单位 #73

实现GAE和GKE k8s cron job的秒级调度单位 #73

Comments

mrdulin commented Apr 8, 2019 • edited

实现GAE和GKE k8s cron job的秒级调度单位

问题

解决方案

GAE cron job

GKE cron job

源码地址

mrdulin commented Apr 8, 2019 •

edited

`GAE` cron job