How to configure `Terraform` to upload zip file to `s3` bucket and then deploy them to lambda - amazon-s3

Question

I use TerraForm as infrastructure framework in my application. Below is the configuration I use to deploy python code to lambda. It does three steps: 1. zip all dependencies and source code in a zip file; 2. upload the zipped file to s3 bucket; 3. deploy to lambda function.
But what happens is the deploy command terraform apply will fail with below error:
Error: Error modifying Lambda Function Code quote-crawler: InvalidParameterValueException: Error occurred while GetObject. S3 Error Code: NoSuchKey. S3 Error Message: The specified key does not exist.
status code: 400, request id: 2db6cb29-8988-474c-8166-f4332d7309de
on config.tf line 48, in resource "aws_lambda_function" "test_lambda":
48: resource "aws_lambda_function" "test_lambda" {
Error: Error modifying Lambda Function Code praw_crawler: InvalidParameterValueException: Error occurred while GetObject. S3 Error Code: NoSuchKey. S3 Error Message: The specified key does not exist.
status code: 400, request id: e01c83cf-40ee-4919-b322-fab84f87d594
on config.tf line 67, in resource "aws_lambda_function" "praw_crawler":
67: resource "aws_lambda_function" "praw_crawler" {
It means the deploy file doesn't exist in s3 bucket. But it success in the second time when I run the command. It seems like a timing issue. After upload the zip file to s3 bucket, the zip file doesn't exist in s3 bucket. That's why the first time deploy failed. But after a few seconds later, the second command finishes successfully and very quick. Is there anything wrong in my configuration file?
The full terraform configuration file can be found: https://github.com/zhaoyi0113/quote-datalake/blob/master/config.tf

Answer 1

You need to add dependency properly to achieve this, Otherwise, it will crash.
First Zip the files
# Zip the Lamda function on the fly
data "archive_file" "source" {
type = "zip"
source_dir = "../lambda-functions/loadbalancer-to-es"
output_path = "../lambda-functions/loadbalancer-to-es.zip"
}
then upload it s3 by specifying it dependency which zip,source = "${data.archive_file.source.output_path}" this will make it dependent on zip
# upload zip to s3 and then update lamda function from s3
resource "aws_s3_bucket_object" "file_upload" {
bucket = "${aws_s3_bucket.bucket.id}"
key = "lambda-functions/loadbalancer-to-es.zip"
source = "${data.archive_file.source.output_path}" # its mean it depended on zip
}
Then you are good to go to deploy Lambda, To make it depened just this line do the magic s3_key = "${aws_s3_bucket_object.file_upload.key}"
resource "aws_lambda_function" "elb_logs_to_elasticsearch" {
function_name = "alb-logs-to-elk"
description = "elb-logs-to-elasticsearch"
s3_bucket = "${var.env_prefix_name}${var.s3_suffix}"
s3_key = "${aws_s3_bucket_object.file_upload.key}" # its mean its depended on upload key
memory_size = 1024
timeout = 900
timeouts {
create = "30m"
}
runtime = "nodejs8.10"
role = "${aws_iam_role.role.arn}"
source_code_hash = "${base64sha256(data.archive_file.source.output_path)}"
handler = "index.handler"
}

Answer 2

You may find that the source_code_hash changes even when the code hasn't changed when using Terraform's archive_file. If this is an issue for you I created a module to fix this: lambda-python-archive.

Answer 3

This is a response to the top answer:
You need to add .output_base64sha256 to the source_code_hash instead of using base64sha256 or else terraform plan never settles with "no changes / up-to-date" message.
For example:
source_code_hash = "${data.archive_file.source.output_bash64sha256}"