sam package too large when deploying custom runtime lambda - swift

Question

So I replicated this project which uses swift as custom lambda runtime using a makefile as build method.
Now I created a AWS CodePipeline that packages my project using CodeBuild using sam package and finally deploys it via CloudFormation.
The codeUri of my lambda is set at the root folder like you see in the repo I linked above. I think that is how it should as I saw that as well in the sam documentation under the custom runtime section. The problem with that is that sam package packages my entire project and lambda is complaining at deploy time that the zip is too large.
How would I set up the makefile as well as the template.yml so that sam package only packages my lambdas?

Answer 1

So I got it to work with a slightly different strategy. This is for anyone who found them selfs in the same situation.
1. Don't use sam to build your lambda functions.
I am running a variety on shell scripts to initiate the swift build in the /scripts folder.
.
├── Package.resolved
├── Package.swift
├── README.md
├── Sources
│ └── YourFirstLambda
│ ├── main.swift
│ └── requirements.txt
├── buildspec.yml
├── samconfig.toml
├── scripts
│ ├── build-and-package-all.sh
│ ├── build-and-package.sh
│ └── package.sh
└── template.yml
build-and-package-all.sh
Start this shell script from inside the scripts folder. You can change this behavior if you change all the dir paths.
This initiate the build-and-packange.sh script for each function defined in the array lambdas.
declare -a lambdas=("YourFirstLambda" "YourSeconLambda")
workspace="$(pwd)/.."
## now loop through the above array
if [ -f /.dockerenv ]; then
# This is executed if run inside docker
echo "I'm inside matrix ;(";
for lambda in "${lambdas[#]}"
do
# Second parameter is wheather we are inside a docker container or not
./build-and-package.sh $lambda "FALSE"
done
else
echo "I'm living in real world!";
for lambda in "${lambdas[#]}"
do
# Second parameter is wheather we are inside a docker container or not
./build-and-package.sh $lambda "TRUE"
done
fi
build-and-package.sh
This script runs
swift build and
package.sh
on a docker container if the build-and-package-all.sh is executed on a bare metal machine. This is useful because you can run this on a machine that does not have swift installed.
On the other hand we will run swift build on bare metal if we are already in a docker container. This might be the case like it was for me when you want to build you functions using AWS CodeBuild. They also use a docker container so there is no need to start a docker container inside a docker container.
set -eu
executable=$1
isBareMetal=$2
workspace="$(pwd)/.."
if [ $isBareMetal == "TRUE" ]; then
echo "-------------------------------------------------------------------------"
echo "building \"$executable\" lambda"
echo "-------------------------------------------------------------------------"
docker run --rm -v "$workspace":/workspace -w /workspace/ codebuild-swift \
bash -cl "swift build --product $executable -c release"
echo "done"
echo "-------------------------------------------------------------------------"
echo "packaging \"$executable\" lambda"
echo "-------------------------------------------------------------------------"
docker run --rm -v "$workspace":/workspace -w /workspace/ codebuild-swift \
bash -cl "sh scripts/package.sh $executable"
echo "done"
else
echo "-------------------------------------------------------------------------"
echo "building \"$executable\" lambda"
echo "-------------------------------------------------------------------------"
cd $workspace
swift build --product $executable -c release
echo "done"
echo "-------------------------------------------------------------------------"
echo "packaging \"$executable\" lambda"
echo "-------------------------------------------------------------------------"
sh $workspace/scripts/package.sh $executable
echo "done"
fi
Finally we package the swift lambda to a .zip.
package.sh
set -eu
executable=$1
target=".build/lambda/$executable"
rm -rf "$target"
mkdir -p "$target"
cp ".build/release/$executable" "$target/"
# add the target deps based on ldd
ldd ".build/release/$executable" | grep swift | cut -d' ' -f3 | xargs cp -Lv -t "$target"
cd "$target"
ln -s "$executable" "bootstrap"
zip --symlinks lambda.zip *
2. Tell sam where to find the zipped lambda
In the template.yml you should have a section that describes your lambda like so:
...
YourLambdaFunction:
Type: AWS::Serverless::Function
Properties:
Timeout: 5
Handler: Provided
Runtime: provided
MemorySize: 128
Description: Test Lambda
Role: !GetAtt Role.Arn
CodeUri: .build/lambda/YourLambdaFunction/lambda.zip
...
You can now use sam build, sam deploy or sam package. Sam will only upload the zipped lambda which should be in the 30Mb range. Probably less for you if you do not have many dependencies.
Side note.
You will need a docker container that has swift installed. My docker image is tagged codebuild-swift and uses the following docker file. If you name your docker image differently you then have to update the build-and-package.sh:
FROM swift:5.2-amazonlinux2
RUN yum -y install \
git \
libuuid-devel \
libicu-devel \
libedit-devel \
libxml2-devel \
sqlite-devel \
python-devel \
ncurses-devel \
curl-devel \
openssl-devel \
tzdata \
libtool \
gcc-c++ \
jq \
tar \
zip \
glibc-static
The shell scripts above are all based from this site:
Getting started with sift AWS Lambda runtime