Can Spring Cloud Function be used with AWS Lambda base image

Question

I try to deploy a function on AWS Lambda by a container image, but I cannot find an example of how to build it with the AWS lambda base image. Does Spring Cloud Function support this deploy method or only support deploying by a jar?

Answer 1

Yes, Spring Cloud Functions will work deployed as a container image. The process will be the same whether you use Spring Cloud Functions or another framework.

Here is an example Dockerfile from lambda-libs GitHub project.

# we'll use Amazon Linux 2 + Corretto 11 as our base
FROM public.ecr.aws/amazoncorretto/amazoncorretto:11 as base

# configure the build environment
FROM base as build
RUN yum install -y maven
WORKDIR /src

# cache and copy dependencies
ADD pom.xml .
RUN mvn dependency:go-offline dependency:copy-dependencies

# compile the function
ADD . .
RUN mvn package 

# copy the function artifact and dependencies onto a clean base
FROM base
WORKDIR /function

COPY --from=build /src/target/dependency/*.jar ./
COPY --from=build /src/target/*.jar ./

# configure the runtime startup as main
ENTRYPOINT [ "/usr/bin/java", "-cp", "./*", "com.amazonaws.services.lambda.runtime.api.client.AWSLambda" ]
# pass the name of the function handler as an argument to the runtime
# in this case it'll be the special Spring Cloud Function handler
CMD [ "example.App::sayHello" ]

Answer 2

I think what you are asking is closely related to AWS Custom Runtime and indeed we do have support and sample for it - https://github.com/spring-cloud/spring-cloud-function/tree/main/spring-cloud-function-samples/function-sample-aws-custom

Answer 3

For both approaches I ran this on Windows WSL2 Ubuntu 20.04 distro. Bash is at the root of the project directory. Here are my two Docker approaches (I was trying to figure this out step by step).

My docker build command

docker build . -t <image name>

You can test your image locally

#note external port = 9000, I left the default spring boot port of 8080
docker run -p 9000:8080 -t <build image name>

Approach A

You can compile your JAR outside of Docker in Maven or Gradle. I then have a docker build file to copy it into the Amazon published Java image. In my case I wanted to deploy my Spring Cloud function using Java 17.

Amazon publishes their ECR base images with higher Java versions (or lower if you need)

https://gallery.ecr.aws/lambda/java (find all Amazon Java images here)

In my Docker build file I did the following - note the handler name is what is for the latest version of Spring Cloud Function.

FROM public.ecr.aws/lambda/java:17

#the docker build command is running at my project root directory
COPY build/libs/<your jar name>.jar ${LAMBDA_TASK_ROOT}/lib/

CMD ["org.springframework.cloud.function.adapter.aws.FunctionInvoker::handleRequest"]

You should be able to run this image locally on your machine and test that this works.

Note: The JAR must be packaged be using the instructions on the Spring Cloud Function page for AWS lambda (https://docs.spring.io/spring-cloud-function/docs/current/reference/html/aws.html). See section Notes on JAR layout.

Approach B

I did write a file later to compile the entire JAR in docker and stick it in an Amazon AWS Java image in a multi-step build. Since Spring Boot comes with a Maven / Gradle wrapper I just used the wrapper instead of trying to install Maven / Gradle in the image itself.

# Use an official OpenJDK runtime as a parent image
FROM openjdk:17-jdk-slim

# Set the working directory to /app
WORKDIR /app

# Copy the application files <this is copies your entire java project dir>
COPY . /app

# Run the Gradle build
RUN ./gradlew clean build

#Copy the resulting JAR file to the ECR image
FROM public.ecr.aws/lambda/java:17

COPY --from=0 /app/build/libs/<INSERT project jar name>.jar ${LAMBDA_TASK_ROOT}/lib/

# Set the handler for AWS Lambda
CMD ["org.springframework.cloud.function.adapter.aws.FunctionInvoker::handleRequest"]