CODEWITHSUNDEEP
amazon-web-servicesaws-lambdaaws-api-gatewayserverless-framework
Jochem Schulenklopper
Jochem Schulenklopper

Reputation: 6934

Specify API Gateway id instead of using 'random' id

With deploying an AWS Lambda function (via Serverless Framework), and exposing it via a HTTPS endpoint in AWS API Gateway... is it possible to construct and set the API Gateway id and thus determine the first part of the HTTP endpoint for that Lambda function?

When deploying an AWS Lambda function and adding a HTTP event, I now get a random id as (the first hostname) in https://klv5e3c8z5.execute-api.eu-west-1.amazonaws.com/v1/fizzbuzz. New/fresh deployments receive new random string, that 10 character id.

Instead of using that, I would like to determine and set that id. (I will make sure that it's sufficiently unique myself, or deal with endpoint naming collisions myself.)

Reason for this is that in a separate Serverless project, I need to use that endpoint (and thus need to know that id). Instead of having it being determined by project 1 and then reading/retrieving that in project 2, I want to construct and set the endpoint in project 1 so that I can use the known endpoint in project 2 as well.

(A suggestion was to use a custom domain as an alternative/alias for that endpoint... but if possible I want don't want to introduce a new component in the mix, and a solution that does not include Cloud-it-might-take-30-minutes-to-create-a-domain-Front is better :-) )

If this isn't possible, I might need to use the approach as described at http://www.goingserverless.com/blog/api-gateway-url, mentioning that the endpoint is being exposed from one project via the CloudFormation stack, to be read from and used in the other project, but that introduces (a little latency and) a dependency in deploying the second project.

Upvotes: 2

Views: 3249

Answers (2)

Jochem Schulenklopper
Jochem Schulenklopper

Reputation: 6934

Question: "is it possible to construct and set the API Gateway id?"

Answer: No (see the other answer to this question).

I was able to get the service endpoint of project 1 into the serverless.yml file of project 2 though, to finally construct the full URL of the service that I needed. I'm sharing this because it's an alternative solution that also works in my case.

In the serverless.yml of project 2, you can refer to the service endpoint of project 1 via service_url: "${cf:<service-name>-<stage>.ServiceEndpoint}". Example: "${cf:my-first-service-dev.ServiceEndpoint}".

CloudFront exposes the ServiceEndpoint that contains the full URL, so including the AWS Gateway REST API id.

More information in Serverless Framework documentation: https://serverless.com/framework/docs/providers/aws/guide/variables/#reference-cloudformation-outputs.

It seems that Serverless Framework is adding this ServiceEndpoint as stack output.

Upvotes: 0

Dunedan
Dunedan

Reputation: 8435

The "first hostname" you want to set is called "REST API id" and is generated by API Gateway when creating the API. The API used to create API's in API Gateway doesn't offer the ability to specify the REST API id, so no, there is no way to specify the id.

The reason for that is probably that these id's are used as part of a public facing domain name. As this domain name doesn't include an identifier for the AWS account it belongs to, the id's have to be globally unique, so AWS generates them to avoid collisions. As AWS puts it (emphasis by me):

For an edge-optimized API, the base URL is of the http[s]://*{restapi-id}*.execute-api.amazonaws.com/stage format, where {restapi-id} is the API's id value generated by API Gateway. You can assign a custom domain name (for example, apis.example.com) as the API's host name and call the API with a base URL of the https://apis.example.com/myApi format.

For the option to create a custom domain name you should consider that there is even some more complexity associated with it, as you must provision a matching SSL-certificate for the domain as well. While you can use ACM for that, there is currently the limitation that SSL-certificates for CloudFront distributions (which edge-optimized API Gateway API's use behind the scenes) need to be issued in us-east-1.

The option you already mentioned to export the API endpoint in the CloudFormation stack as output value and use that exported value in your other stack would work well. As you noted that'd create a dependency between the two stacks, so once you deployed project 2, which uses the output value from project 1, you can only delete the CloudFormation stack for project 1 after the project 2 stack is either deleted or updated to not use the exported value anymore. That can be a feature, but from your description it sounds like it wouldn't for your use case.

Something similar to exported stack output values would be to use some shared storage instead of making use of CloudFormation's exported output values features. What comes to mind here is the SSM Parameter Store, which offers some integration into CloudFormation. The integration makes it easy to read a parameter from the SSM Parameter Store in the stack of project 2. For writing the value to the Parameter Store in project 1 you'd need to use a custom resource in your CloudFormation template. There is at least one sample implementation for that available on Github.

As you can see there are multiple options available to solve your problem. Which one to choose depends on your projects needs.

Upvotes: 2

Related Questions