AWS Step Functions 将多个 AWS 服务协调为无服务器工作流,以便可以快速构建和更新应用程序。使用 Step Functions,可以设计和运行将AWS Lambda 和 Amazon ECS 等服务整合到功能丰富的应用程序中的工作流。工作流由一系列步骤组成,一个步骤的输出充当下一个步骤的输入。 使用 Step Functions,应用程序开发更简单、更直观,因为它将工作流转换为易于理解、易于向其他人说明且易于更改的状态机示意图。可以监控执行的每个步骤,这意味着可以快速识别并解决问题。Step Functions 可以自动触发和跟踪各个步骤,并在出现错误时重试,因此的应用程序能够按照预期顺序执行。
demo架构图
AWS step function 状态机的工作流程
工作流程如图:
1.调用 Input Lottery Winners 函数,传入 num_of_winners,进入第二步。
2.Random Select Winners 根据 Input Lottery Winners 的输出(body)调用 Random Select Winners,生成两个获奖号码,进入第三步。
3.Validate Winners 根据第二步输出查询 Winners 表判断是否重复,重复则传出 status:1,否则传出 status:0。
4.Is Winner In Past Draw 接受第三步 status 并判断,当 status 为1,则重新调用 Random Select Winners 进入第二步,当 status 为0,则调用 Notify Winners 和 Record Winner Queue,给 SNS topic 发送通知,把获奖者写入 Winner 表。
在整个工作流程中,当 Catch 接收到错误,都直接进入 Failed 步骤,输出异常并中断 step function。
- 创建步骤
IAM角色创建
执行 lambda 的角色需要以下策略: AmazonDynamoDBFullAccess
AWSLambdaBasicExecutionRole
AmazonSNSFullAccess
AWSStepFunctionsFullAccess
在 AWS IAM 控制台中创建角色:
大概如下图
lambda 创建
Input Lottery Winners
为了实现Step Functions状态机流转下的任务,我们这次实现会用到AWS Lambda作为我们业务的实现环境 1.进入AWS控制台,选择服务然后输入Lambda进入AWS Lambda控制台
2.选择创建函数,然后选择从头开始创作来自定义我们的实验程序
3.首先我们需要创建状态机中的第一个状态任务Input Lottery Winners,输入函数名称Lottery-InputWinners来定义幸运儿的数量。运行语言选择Python 3.7。同时需要选择函数执行的权限, 这里我们选择使用现用角色,选择我们之前创建的IAM角色
4.点击创建函数
5.在函数代码栏目下输入如下代码块,修改代码中的 region_name 为当前使用的 region
6.创建函数时复制页面右上角的ARN,后面创建状态机需要
import json class CustomError(Exception): pass def lambda_handler(event, context): num_of_winners = event['input'] # Trigger the Failed process if 'exception' in event: raise CustomError("An error occurred!!") return { "body": { "num_of_winners": num_of_winners } }
接下来我们还需要创建另外三个需要定义的状态机业务逻辑,创建过程和上面的Lottery-InputWinners一致,下面是具体的状态机AWS
Lambda代码块
Lottery-RandomSelectWinners
import json import boto3 from random import randint from boto3.dynamodb.conditions import Key, Attr TOTAL_NUM = 10 def lambda_handler(event, context): # variables num_of_winners = event['num_of_winners'] # query in dynamodb dynamodb = boto3.resource('dynamodb', region_name='') table = dynamodb.Table('Lottery-Employee') # random select the winners, if has duplicate value, re-run the process while True: lottery_serials = [randint(1,TOTAL_NUM) for i in range(num_of_winners)] if len(lottery_serials) == len(set(lottery_serials)): break # retrieve the employee details from dynamodb results = [table.query(KeyConditionExpression=Key('lottery_serial').eq(serial), IndexName='lottery_serial-index') for serial in lottery_serials] # format results winner_details = [result['Items'][0] for result in results] return { "body": { "num_of_winners": num_of_winners, "winner_details": winner_details } }
Lottery-ValidateWinners
import json
import boto3
from boto3.dynamodb.conditions import Key, Attr
def lambda_handler(event, context):
# variables
num_of_winners = event['num_of_winners']
winner_details = event['winner_details']
# query in dynamodb
dynamodb = boto3.resource('dynamodb', region_name='')
table = dynamodb.Table('Lottery-Winners')
# valiate whether the winner has already been selected in the past draw
winners_employee_id = [winner['employee_id'] for winner in winner_details]
results = [table.query(KeyConditionExpression=Key('employee_id').eq(employee_id)) for employee_id in winners_employee_id]
output = [result['Items'] for result in results if result['Count'] > 0]
# if winner is in the past draw, return 0 else return 1
has_winner_in_queue = 1 if len(output) > 0 else 0
# format the winner details in sns
winner_names = [winner['employee_name'] for winner in winner_details]
name_s = ""
for name in winner_names:
name_s += name
name_s += " "
return {
"body": {
"num_of_winners": num_of_winners,
"winner_details": winner_details
},
"status": has_winner_in_queue,
"sns": "Congrats! [{}] You have selected as the Lucky Champ!".format(name_s.strip())
}
Lottery-RecordWinners
import json import boto3 from boto3.dynamodb.conditions import Key, Attr def lambda_handler(event, context): # variables winner_details = event['winner_details'] # retrieve the winners' employee id employee_ids = [winner['employee_id'] for winner in winner_details] # save the records in dynamodb dynamodb = boto3.resource('dynamodb', region_name='') table = dynamodb.Table('Lottery-Winners') for employee_id in employee_ids: table.put_item(Item={ 'employee_id': employee_id }) return { "body": { "winners": winner_details }, "status_code": "SUCCESS" }
创建AWS SNS 通知服务
1.进入AWS控制台,在服务中搜索SNS
2.在SNS控制面板中,选择主题, 然后选择创建主题
3.在创建新主题弹框中,输入
主题名称: Lottery-Notification
显示名称: Lottery
1.创建主题后,会进入主题详细信息页面,这时候我们需要创建订阅来对接我们的消息服务,例如邮件服务(本次实验使用邮件服务来作为消息服务)
2.点击创建订阅, 在弹框中选择
协议: Email
终端节点: <填入自己的邮箱地址>
1.点击请求确认, 然后到上面填写的邮箱地址中确认收到信息,表示确认该邮箱可以接收来自AWS SNS该主题的通知消息
2.复制主题详细页面的主题ARN,之后替换Step Functions状态机下的Notification:ARN
创建Amazon Dynamodb 服务
本次实验需要创建两张Dynamodb表来记录员工信息和幸运儿信息。使用Dynamodb能更快地通过托管的方式记录数据同时免去数据库运维的压力。
1.进入AWS控制台,在服务中搜索Dynamodb
2.在左侧控制栏中选在表, 然后在主页面中选择创建表
3.在创建Dynamodb表中,填入如下信息
·表名称:Lottery-Winners
·主键:employee_id
4.表设置中确认勾选使用默认设置,点击创建
5.同样的设置步骤,点击创建表,在创建Dynamodb表中,填入如下信息
·表名称:Lottery-Employee主键:employee_id
1.表设置中确认勾选使用默认设置,点击创建
2.等待表创建完成后, 通过本附件中的request-items.json文件导入数据到Lottery-Employee
$ aws dynamodb batch-write-item --request-items file://request-items.json
1.选择表Lottery-Employee Tab页面中的索引, 点击创建索引
主键:lottery_serial, 字段类型选择数字
索引名称:lottery_serial-index
创建AWS Step Functions 状态机
1.进入AWS控制台,在服务中搜索Step Functions
2.进入Step Functions服务后,点击左侧的活动栏,并点击状态机
3.进入状态机主页面后,选择创建状态机
4.在定义状态机栏目下,选择默认使用代码段创作。同时在详细信息栏目输入状态机名称Lottery
5.在状态机定义栏目下,复制如下状态机定义文件,通过Amazon States Language来定义状态机的状态流转
{
"Comment": "A simple AWS Step Functions state machine that simulates the lottery session",
"StartAt": "Input Lottery Winners",
"States": {
"Input Lottery Winners": {
"Type": "Task",
"Resource": "<InputWinners:ARN>",
"ResultPath": "$",
"Catch": [
{
"ErrorEquals": [ "CustomError" ],
"Next": "Failed"
},
{
"ErrorEquals": [ "States.ALL" ],
"Next": "Failed"
}
],
"Next": "Random Select Winners"
},
"Random Select Winners": {
"Type": "Task",
"InputPath": "$.body",
"Resource": "<RandomSelectWinners:ARN>",
"Catch": [
{
"ErrorEquals": [ "States.ALL" ],
"Next": "Failed"
}
],
"Retry": [
{
"ErrorEquals": [ "States.ALL"],
"IntervalSeconds": 1,
"MaxAttempts": 2
}
],
"Next": "Validate Winners"
},
"Validate Winners": {
"Type": "Task",
"InputPath": "$.body",
"Resource": "<ValidateWinners:ARN>",
"Catch": [
{
"ErrorEquals": [ "States.ALL" ],
"Next": "Failed"
}
],
"Retry": [
{
"ErrorEquals": [ "States.ALL"],
"IntervalSeconds": 1,
"MaxAttempts": 2
}
],
"Next": "Is Winner In Past Draw"
},
"Is Winner In Past Draw": {
"Type" : "Choice",
"Choices": [
{
"Variable": "$.status",
"NumericEquals": 0,
"Next": "Send SNS and Record In Dynamodb"
},
{
"Variable": "$.status",
"NumericEquals": 1,
"Next": "Random Select Winners"
}
]
},
"Send SNS and Record In Dynamodb": {
"Type": "Parallel",
"End": true,
"Catch": [
{
"ErrorEquals": [ "States.ALL" ],
"Next": "Failed"
}
],
"Retry": [
{
"ErrorEquals": [ "States.ALL"],
"IntervalSeconds": 1,
"MaxAttempts": 2
}
],
"Branches": [
{
"StartAt": "Notify Winners",
"States": {
"Notify Winners": {
"Type": "Task",
"Resource": "arn:aws:states:::sns:publish",
"Parameters": {
"TopicArn": "<Notification:ARN>",
"Message.$": "$.sns"
},
"End": true
}
}
},
{
"StartAt": "Record Winner Queue",
"States": {
"Record Winner Queue": {
"Type": "Task",
"InputPath": "$.body",
"Resource":
"<RecordWinners:ARN>",
"TimeoutSeconds": 300,
"End": true
}
}
}
]
},
"Failed": {
"Type": "Fail"
}
}
}
1.在状态机定义栏目的右侧,点击刷新按钮,可以看到状态机流转的流程图。使用之前的 lambda ARN(4个),SNS topic ARN(1个),对应替换状态机 json 文件中的InputWinners:ARN,RandomSelectWinners:ARN,ValidateWinners:ARN,RecordWinners:ARN,Notification:ARN,点击下一步。
2.在配置设置下,选择为我创建IAM角色, 输入自定义的IAM角色名称MyStepFunctionsExecutionRole,并且附加AmazonSNSFullAccess权限
3.点击创建状态机完成创建过程
- 执行 Step Function 状态机 1.入AWS控制台,在服务中搜索Step Functions
2.进入之前创建的状态机Lottery
3.点击启动执行
4.在弹框中填入输入的json文本,这里的input代表在本次实验中需要抽取的获奖人数
{ "input": 2 } 1.点击启动执行
- 实验结果 1.Dynamodb表中Lottery-Winners记录获奖者
2.邮件会收取到幸运儿的信息
- 总结 在step function控制台中可以详细观察到整个函数链的执行过程,点击可视工作流中的函数名可以详细查看各个函数的输入输出。
当函数出错时,也能直观得在可视化界面看到,并且在输出中查看错误日志