Skip to content

Latest commit

 

History

History
46 lines (32 loc) · 1.69 KB

Compilation-Prerequisites.md

File metadata and controls

46 lines (32 loc) · 1.69 KB

Compilation Prerequisites

Overview

When you create a Lambda you can choose either x86_64 or arm64. Both Python and Node are interpreted from source code (or bytecode) so usually either will work. But Python can call out to C and Rust binary code. If your Python code uses these, it must not only match the architecture, it also must have the same exe format, so Mac ARM (Mach-O) won't work as a Linux based (ELF) Lambda.

We have one Layer that does that: package_py. This called by two Lambdas: etl_task_file_copy and etl_task_sftp. Deploying these from our Macs takes a little more work. We have two solutions, both using Docker.

Deploying from Mac M-series

All Lambdas and Layers can be built using one of these two combinations:

  • Option 1, deploy locally

    • In make_variables, set build_mode="sam" and architecture="x86_64"
    • Prerequisites:
      • Nodejs 20.x
      • Python 3.12
      • Terraform
      • Docker
      • AWS SAM
      • git, make and zip

  • Option 2, using a Docker image

    • In make_variables, set build_mode="std" and architecture="arm64"

    • Prerequisites:

      • Docker

    • To install on Docker, see here.

Deploying from Linux (and presumably Windows WSL)

  • In make_variables, set build_mode="std" and architecture to the architecture of your computer.
  • Prerequisites:
    • Nodejs 20.x
    • Python 3.12
    • Terraform
    • git, make and zip

Docker should work anywhere by setting 'architecture=' to the architecture of your computer.

Combinations of build_mode and architecture

With the exception of package_py, using build_mode="std" with either architecture ("x86_64" or "arm64") will work inside or out of Docker, for all other Lambdas and Layers.