serverless-rust

A ⚡ Serverless framework ⚡ plugin for Rustlang applications

📦 Install

Install the plugin inside your serverless project with npm.

``sh
$ npm i -D serverless-rust
`
💡The -D flag adds it to your development dependencies in npm speak

💡 This plugin assumes you are building Rustlang lambdas targeting the AWS Lambda "provided" runtime. The AWS Lambda Rust Runtime makes this easy.

Add the following to your serverless project's serverless.yml file

`yaml
service: demo
provider:
name: aws
runtime: rust
plugins:
# this registers the plugin
# with serverless
- serverless-rust


creates one artifact for each function


package:
individually: true
functions:
test:
# handler value syntax is {cargo-package-name}.{bin-name}
# or {cargo-package-name} for short when you are building a
# default bin for a given package.
handler: your-cargo-package-name
events:
- http:
path: /test
method: GET
`

> 💡 The Rust Lambda runtime requires a binary named bootstrap. This plugin renames the binary cargo builds to bootstrap for you. You do not need to do this manually in your Cargo.toml configuration file.

The default behavior is to build your lambda inside a docker container. Make sure you have a docker daemon running if you are not opting into the dockerless mode.

🖍️ customize

You can optionally adjust the default settings of the dockerized build env using
a custom section of your serverless.yaml configuration

`yaml
custom:
# this section customizes of the default
# serverless-rust plugin settings
rust:
# flags passed to cargo
cargoFlags: '--features enable-awesome'
# custom docker tag
dockerTag: 'some-custom-tag'
# custom docker image
dockerImage: 'dockerUser/dockerRepo'
`

$3

While it's useful to have a build environment that matches your deployment
environment, dockerized builds come with some notable tradeoffs.

The external dependency on docker itself often causes friction as an added dependency to your build.

Depending on a docker image limits which versions of rust you can build with. The default docker image tracks stable rust. Some users might wish to try unstable versions of rust before they stabilize. Local builds enable that.

If you wish to build lambda's locally, use the dockerless configuration setting.

`diff
custom:
# this section allows for customization of the default
# serverless-rust plugin settings
rust:
# flags passed to cargo
cargoFlags: '--features enable-awesome'
# experimental! when set to true, artifacts are built locally outside of docker
+ dockerless: true

# when using local builds (dockerless), optionally provide a different target and linker for the compiler
# for example, allow local running on ARM macs
target: aarch64-apple-darwin
linker: clang
`

The following assumes that you have not specified a different target or linker. If you do, make sure have that you have installed the specified target (via rustup) and linker.

This will build and link your lambda as a static binary outside a container that can be deployed in to the lambda execution environment using MUSL. The aim is that in future releases, this might become the default behavior.

In order to use this mode its expected that you install the x86_64-unknown-linux-musl target on all platforms locally with

`sh
$ rustup target add x86_64-unknown-linux-musl
`

On linux platforms, you will need to install musl-tools

`sh
$ sudo apt-get update && sudo apt-get install -y musl-tools
`

On Mac OSX, you will need to install a MUSL cross compilation toolchain

`sh
$ brew install filosottile/musl-cross/musl-cross
`

Using MUSL comes with some other notable tradeoffs. One of which is complications that arise when depending on dynamically linked dependencies.

* With OpenSSL bindings which you can safely replace is with rustls or vendor it
* Other limitations are noted here.

If you find other MUSL specific issues, please report them by opening an issue.

$3

If your serverless project contains multiple functions, you may sometimes
need to customize the options above at the function level. You can do this
by defining a rust key with the same options inline in your function
specification.

`yaml
functions:
test:
rust:
# function specific flags passed to cargo
cargoFlags: '--features enable-awesome'
# handler value syntax is {cargo-package-name}.{bin-name}
# or {cargo-package-name} for short when you are building a
# default bin for a given package.
handler: your-cargo-package-name
events:
- http:
path: /test
method: GET
`

🤸 usage

Every serverless workflow command should work out of the box.

$3

`sh
$ npx serverless invoke local -f hello -d '{"hello":"world"}'
`

$3

`sh
$ npx serverless deploy
`

$3

`sh
$ npx serverless invoke -f hello -d '{"hello":"world"}'
`

$3

`sh
$ npx serverless logs -f hello
``

🏗️ serverless templates

$3

* a minimal echo application - https://github.com/softprops/serverless-aws-rust
* a minimal http application - https://github.com/softprops/serverless-aws-rust-http
* a minimal multi-function application - https://github.com/softprops/serverless-aws-rust-multi
* a minimal apigateway websocket application - https://github.com/softprops/serverless-aws-rust-websockets
* a minimal kinesis application - https://github.com/softprops/serverless-aws-rust-kinesis

$3

Older versions targeted the python 3.6 AWS Lambda runtime and rust crowbar and lando applications

* lando api gateway application - https://github.com/softprops/serverless-lando
* multi function lando api gateway application - https://github.com/softprops/serverless-multi-lando
* crowbar cloudwatch scheduled lambda application - https://github.com/softprops/serverless-crowbar

Doug Tangren (softprops) 2018-2019