crambda

Crystal AWS Lambda custom runtime

Installation

  1. Add the dependency to your shard.yml:
dependencies:
  crambda:
    github: lambci/crambda

targets:
  bootstrap:
    main: src/main.cr
  1. Run shards install

Usage

(Assuming this is your src/main.cr or similar from above)

require "json"
require "crambda"

def handler(event : JSON::Any, context : Crambda::Context)
  pp context
  JSON.parse("[1, 2]")
end

Crambda.run_handler(->handler(JSON::Any, Crambda::Context))

Where Crambda.run_handler expects a handler that takes a JSON::Any event and a Context, and returns a JSON::Any response:

def self.run_handler(handler : Proc(JSON::Any, Context, JSON::Any))
  # ...
end

And Context is a class that looks like this:

class Context
  getter function_name : String
  getter function_version : String
  getter memory_limit_in_mb : UInt32
  getter log_group_name : String
  getter log_stream_name : String
  getter aws_request_id : String
  getter invoked_function_arn : String
  getter deadline_ms : Int64
  getter identity : JSON::Any
  getter client_context : JSON::Any

  def get_remaining_time_in_millis
    # ...
  end
nd

Compiling and uploading to AWS Lambda

Static binary (easy, but larger and slower)

Creating a static binary is the easiest method, but will be larger and slower than using dynamic libraries.

If you're on Linux already, you can do:

shards install

shards build --release --no-debug --static

strip bin/bootstrap # optional, to reduce size

Then package up bootstrap at the top level of a zipfile:

cd bin
zip lambda.zip bootstrap

And upload lambda.zip to your custom runtime AWS Lambda.

If you're not on Linux, you can run the install step locally (if you have crystal – brew install crystal on MacOS), and then compile in a docker container:

shards install

docker run --rm -v "$PWD":/app -w /app crystallang/crystal sh -c \
  'shards build --release --no-debug --static && strip bin/bootstrap'

(then zip up your bootstrap executable as above)

Dynamically linked binary (more steps, but smaller and faster)

The only libs that need to be statically linked in your binary (ie, that don't exist on AWS Lambda) are libevent, libgc and libcrystal. By default, crystal statically links the last two anyway, but libevent doesn't exist on Lambda, so either needs to be uploaded as a separate .so alongside your bootstrap, or compiled in.

The most straightforward way to link these libs into your binary is to use the ones supplied in the ext directory in crambda, as follows:

First build a cross-compiled version of your Lambda function (you can do this on any machine that has crystal, including MacOS):

shards install

PKG_CONFIG_PATH=lib/crambda/ext crystal build src/main.cr -o bin/bootstrap \
  --release --no-debug --cross-compile --target x86_64-unknown-linux-gnu

# Ignore the command it outputs – it needs to be modified slightly as below

This will create a bin/bootstrap.o object file that you can link in a Lambda-like environment – eg, a machine running Amazon Linux, or in a Docker container:

docker run --rm -v "$PWD":/var/task lambci/lambda:build-provided cc bin/bootstrap.o -o bin/bootstrap -s \
  -rdynamic -lz -lssl -lcrypto -lpcre -lm -lgc -lpthread -lcrystal -levent -lrt -ldl -Llib/crambda/ext

This will place the bootstrap binary in bin where you can zip it up and upload to Lambda as shown above in the static binary instructions.

If you're using a different version of crystal from the one supplied in ext (currently 0.28.0), then you'll need to replace -lcrystal with the path to the version of libcrystal that matches your environment.