Skip to main content

Containers

Use containers in your Lambda functions.

Overview​

There may be instances when your code exceeds the 250MB Lambda limit. Examples could be video processing or ML tasks with large dependencies. In such cases, containers can be a viable solution. Lambda container functions allows for a maximum size of 10GB.

Get started​

Let's look at an example where we invoke a Lambda container function through a cron job.

Start by creating a new SST app by running the following command in your terminal.

npx create-sst@latest

Add the construct​

Import the Cron construct, and it to your stack.

stacks/MyStack.ts
import { Cron } from "sst/constructs";

new Cron(stack, "cron", {
  schedule: "rate(1 minute)",
  job: {
    function: {
      runtime: "container",
      handler: "packages/functions/src/cron",
    }
  }
});

The cron job will run every minute and points to a container function.

Add the handler​

Now let's implement the function that'll be invoked. Create a file in packages/functions/src/cron/handler.py.

packages/functions/src/cron/handler.py
import numpy

def main(event, context):
    n = 16
    print(f"Square root of {n} is {numpy.sqrt(n)}")
    return "ok"

This function prints a message and calculates the square root of 16 using numpy.

Create a requirements.txt file and listing numpy in it.

numpy

Add the Dockerfile​

Next, let's add a Dockerfile to package our function into a container.

packages/functions/src/cron/Dockerfile
# Start from AWS Python 3.8 base image
FROM public.ecr.aws/lambda/python:3.8

# Install the dependencies
COPY requirements.txt .
RUN pip3 install -r requirements.txt --target "${LAMBDA_TASK_ROOT}"

# Copy our function code
COPY handler.py ${LAMBDA_TASK_ROOT}

# Set the handler function
CMD [ "handler.main" ]

If you run sst dev, you will notice Square root of 16 is 4.0 printed out every minute in the terminal.

FAQ​

Here are some frequently asked questions about Container Functions.

When should I use container function?​

Since container functions have a much longer cold start than normal Lambda functions, it is recommended to use them for async tasks. For example, as event subscribers and cron jobs.

Why did the function timeout locally?​

When running sst dev, the image for a container function is built on the first function invocation. If there are uncached layers that need building, docker build may take longer to run. It is recommended to use the --increase-timeout option when running sst dev.