Deploying changes to individual Lambdas using Serverless Framework

I have a serverless project that deploys 2 Lambdas in the same stack:

service: example-apis2

provider:
  name: aws
  memorySize: 512
  region: us-west-1
  apiGateway:
    restApiId: ${env:APIGWID} #  API Gateway to add this api to
    restApiRootResourceId: ${env:RESOURCEID}
functions:
  example2:
    handler: index2.handler
    layers:
      - arn:aws:lambda:us-west-1:[myaccountid]:layer:example-layer:1
    events:
      - http:
          path: api2
          method: get
  example3:
    handler: index3.handler
    layers:
      - arn:aws:lambda:us-west-1:[myaccountid]:layer:example-layer:1
    events:
      - http:
          path: api3
          method: get

After first deploy, if I do

aws lambda get-function --function-name example-apis2-dev-example2 
...
"LastModified": "2021-10-27T06:30:31.002+0000",

and

$ aws lambda get-function --function-name example-apis2-dev-example3
...
"LastModified": "2021-10-27T06:30:31.987+0000",

Now if I make a code change only to the example 3 Lambda and redeploy only that function with:

serverless deploy function -f example-apis2-dev-example3

… example2 has not been modified since the first deploy (same timestamp as the original deploy):

$ aws lambda get-function --function-name example-apis2-dev-example2
...
"LastModified": "2021-10-27T06:30:31.002+0000",,

and only example3 shows it was updated/redeployed:

$ aws lambda get-function --function-name example-apis2-dev-example3
...
"LastModified": "2021-10-27T06:33:17.736+0000",

serverless deploy : deploys the whole stack (but if nothing has changed there is no update)

serverless deploy function -f functioname: updates just the code on that one Lambda (and updates in a couple of seconds vs several seconds for updating the whole stack).

This is described in this article here.

Deploying multiple Serverless Framework apis to the same AWS API Gateway

By default, each Serverless project you deploy will create a new API Gateway. In most cases this works fine, but for larger projects you may need to split your apis across multiple smaller Serverless projects, each with their own serverless.yml that can be deployed independently.

The Serverless docs describe how to do this here. In each additional Serverless project where you want to add additional apis to an existing API Gateway, you need to specify 2 additional properties in your Serverless.yml, apiGateway and restApiRootResourceId:

provider:
  name: aws
  apiGateway:
    restApiId: xxxxxxxxxx # REST API resource ID. Default is generated by the framework
    restApiRootResourceId: xxxxxxxxxx # Root resource, represent as / path

apiGateway – this is the 11 character id for your API Gateway that you want to add resources to. You can get this from the console and it’s the prefix in your api gw url, e.g. https://aaaaaaaaaaa.execute-api.us-west-1.amazonaws.com/dev

The id for the root resource is where in your api path structure you want to add your new resource to, either the id of the root / or one of the existing paths beneath the root.

This id value I don’t think is visible in the console, but you can get it a list of all the resources in your API Gateway including the ids of each of the existing resources, with:

aws apigateway get-resources --rest-api-id aaaaaaaaaaa --region us-west-2

It will give a response that looks like:

{
    "items": [
        {
            "id": "bbbbbb",
            "parentId": "aaaaaaaaaa",
            "pathPart": "example1",
            "path": "/example1",
            "resourceMethods": {
                "GET": {}
            }
        },
        {
            "id": "aaaaaaaaaa",
            "path": "/"
        }
    ]
}

In this example I have a root / with id = aaaaaaaaa and a resource bbbbbb for /example1.

In this case if I pass aaaaaaaaaa as the value for restApiRootResourceId then my new resource will be added to /, or passing bbbbbb it will be added as a resource under /example1

Enabling AWS API Gateway CloudWatch logging

I deployed a new Lamdba with API Gateway, and when I tried turning on the CloudWatch logging for this API Gateway from the console:

… I got this error that I haven’t seen before:

Turns out per the steps on this page, you need to create an IAM role with API Gateway as the Trusted Entity, and attach the managed policy ‘AmazonAPIGatewayPushToCloudWatchLogs’ :

Add the ARN for the role you created to the Settings for the API you are working with here:

Done!

Running aitextgen model training in a Docker container

I’m setting up an approach to run text generation model training jobs on demand with aitextgen, and the first approach I’m looking at is to run the training in a Docker container. Later I may move this to an AWS service like ECS, but this is my first step.

I’ve built a Docker image with the following dockerfile:

FROM amazonlinux
RUN yum update -y
RUN yum install -y python3
RUN pip3 install aitextgen
ADD source-file-for-fine-tuning.txt .
ADD generate.py .
ADD train.py .

.. and then built my image with:

docker build -t aitextgen .

I then run a container passing in the cmd I want to run, in this case ‘python3 train.py’:

docker run --volume /data/trained_model:/trained_model:rw -d aitextgen sh -c "cd / && python3 train.py && mv aitextgen.tokenizer.json /trained_model"

I’m also attaching a bind point where the model output is being written to during the run, and -d to run the container in the background. The last step in the run command copies the token file to the mounted EBS volume so it can be reused by the generation.

To generate text from the model, run:

docker run --volume /data/trained_model:/trained_model:rw -d aitextgen sh -c "cd / && python3 generate.py"