问题
In AWS CloudFormation, you can define a stack of services with a JSON/YAML template. In particular, I am using Redshift and Aurora clusters.
The template allows you to define administrative properties of the cluster, but there doesn't seem to be a clear way to define tables, views, and indexes in the cluster such that when I push updates of stack, the DB's automatically update the schema as needed, similar to the way it manages updates to full stacks.
Is there a way to manage the schema of Redshift and Aurora databases in Cloud Formation?
回答1:
Amazon CloudFormation can be used to create AWS resources. Think of it as calling the standard AWS APIs on your behalf.
Where the 'contents' of a resource can be controlled by an AWS API call, CloudFormation can also manage it. For example, creating a table in Amazon DynamoDB, a bucket in Amazon S3 and a stream in Amazon Kinesis. These are examples of services that are region-wide and fully provided by AWS.
In contrast, where a service must be deployed within an Availability Zone, CloudFormation has no access. Examples are: Amazon EC2 instances, Amazon RDS instances, Amazon Redshift clusters, Amazon EMR clusters and Amazon Elasticsearch clusters. They all run on top of a virtual machine and the 'contents' are not accessible by AWS API calls.
The only way to perform automatic actions in such systems is to call them directly. For example, establishing a JDBC connection to a database and then passing some SQL commands. This could be done from an application running on an Amazon EC2 instance, from an AWS Lambda function or from any computer connected to the Internet.
Amazon CloudFormation can trigger an AWS Lambda-backed Custom Resources. This Lambda function could then perform practically any function, but you would be responsible for coding this function.
回答2:
Work in Progess!! I'll update once I am sure that it is working!
So crawling documentation, Github and an hour on chat with AWS yields that I might need to create a parent and child stack (example templates). The parent stack would contain a SNS Topic, SNS Subscription and Lambda Function that monitor the child stack.
A parent stack might look like,
AWSTemplateFormatVersion: '2010-09-09'
Resources:
StackEventTopic:
Type: AWS::SNS::Topic
StackEventEmailSubscription:
Type: AWS::SNS::Subscription
Properties:
Protocol: email
Endpoint: <your email>@gmail.com
TopicArn:
Ref: StackEventTopic
StackEventLambdaSubscription:
Type: AWS::SNS::Subscription
Properties:
Protocol: lambda
Endpoint:
Ref: StackEventHandler
TopicArn:
Ref: StackEventTopic
StackEventHandler:
Type: AWS::Lambda::Function
Properties:
... configure your Lambda ...
ChildStack:
Type: AWS::CloudFormation::Stack
Properties:
NotificationARNs:
Ref: StackEventTopic
TemplateURL: "s3 url for template YAML"
来源：https://stackoverflow.com/questions/45553456/defining-tables-views-and-indexes-in-aws-cloudformation