Pushing the custom Python algorithm container image to an Amazon ECR repository

In the previous recipe, we have prepared and built the custom container image using the docker build command. In this recipe, we will push the custom container image to an Amazon ECR repository. If this is your first time hearing about Amazon ECR, it is simply a fully managed container registry that helps us manage our container images.

After pushing the container image to an Amazon ECR repository, we can use this image for training and deployment in the Using the custom Python algorithm container image for training and inference with Amazon SageMaker Local Mode recipe.

Getting ready

Here are the prerequisites for this recipe:

• This recipe continues from the Building and testing the custom Python algorithm container image recipe.
• You will need the necessary permissions to manage the Amazon ECR resources if you're using an AWS IAM user with a custom URL.

How to do it…

The initial steps in this recipe focus on creating the ECR repository. Let's get started:

1. Use the search bar in the AWS Console to navigate to the Elastic Container Registry console. Click Elastic Container Registry when it appears in the search results:
   

   Figure 2.62 – Navigating to the ECR console

   As you can see, we can use the search bar to quickly navigate to the Elastic Container Registry service. If we type in ecr, the Elastic Container Registry service in the search results may come up in third or fourth place.

2. Click the Create repository button:
   

   Figure 2.63 – Create repository button

   Here, the Create repository button is at the top right of the screen.

3. In the Create repository form, specify a Repository name. Use the value of $IMAGE_NAME from the Building and testing the custom Python algorithm container image recipe. In this case, we will use chap02_python:
   

   Figure 2.64 – Create repository form

   Here, we have the Create repository form. For Visibility settings, we will choose Private and set the Tag immutability configuration to Disabled.

4. Scroll down until you see the Create repository button. Leave the other configuration settings as-is and click Create repository:
   

   Figure 2.65 – Create repository button

   As we can see, the Create repository button is at the bottom of the page.

5. Click chap02_python:
   

   Figure 2.66 – Link to the ECR repository page

   Here, we have a link under the Repository name column. Clicking this link should redirect us to the repository's details page.

6. Click View push commands:
   

   Figure 2.67 – View push commands button (upper right)

   As we can see, the View push commands button is at the top right of the page, beside the Edit button.

7. You may optionally copy the first command, aws ecr get-login-password …, from the dialog box.
   

   Figure 2.68 – Push commands dialog box

   Here, we can see multiple commands that we can use. We will only need the first one (aws ecr get-login-password …). Click the icon with two overlapping boxes on the right-hand side of the code box to copy the entire line to the clipboard.

8. Navigate back to the AWS Cloud9 environment IDE and create a new Terminal. You may also reuse an existing one:
   

   Figure 2.69 – New Terminal

   The preceding screenshot shows us how to create a new Terminal. Click the green plus button and then select New Terminal from the list of options. Note that the green plus button is directly under the Editor pane.

9. Navigate to the ml-python directory:

   cd /home/ubuntu/environment/opt/ml-python

10. Get the account ID using the following commands:

    ACCOUNT_ID=$(aws sts get-caller-identity | jq -r ".Account")
    echo $ACCOUNT_ID

11. Specify the IMAGE_URI value and use the ECR repository name we specified while creating the repository in this recipe. In this case, we will run IMAGE_URI="chap02_python":

    IMAGE_URI="<insert ECR Repository URI>"
    TAG="1"

12. Authenticate with Amazon ECR so that we can push our Docker container image to an Amazon ECR repository in our account later:

    aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin $ACCOUNT_ID.dkr.ecr.us-east-1.amazonaws.com

    Important note

    Note that we have assumed that our repository is in the us-east-1 region. Feel free to modify the region in the command if needed. This applies to all the commands in this chapter.

13. Use the docker tag command:

    docker tag $IMAGE_URI:$TAG $ACCOUNT_ID.dkr.ecr.us-east-1.amazonaws.com/$IMAGE_URI:$TAG

14. Push the image to the Amazon ECR repository using the docker push command:

    docker push $ACCOUNT_ID.dkr.ecr.us-east-1.amazonaws.com/$IMAGE_URI:$TAG

    At this point, our custom container image should now be successfully pushed into the ECR repository.

Now that we have completed this recipe, we can proceed with using this custom container image for training and inference with SageMaker in the next recipe. But before that, let's see how this works!

How it works…

In the previous recipe, we used the docker build command to prepare the custom container image. In this recipe, we created an Amazon ECR repository and pushed our custom container image to the repository. With Amazon ECR, we can store, manage, share, and run custom container images anywhere. This includes using these container images in SageMaker during training and deployment.

When pushing the custom container image to the Amazon ECR repository, we need the account ID, region, repository name, and tag. Once we have these, the docker push command will look something like this:

docker push <ACCOUNT_ID>.dkr.ecr.<REGION>.amazonaws.com/<REPOSITORY NAME>:<TAG>

When working with container image versions, make sure to change the version number every time you modify this Dockerfile and push a new version to the ECR repository. This will be helpful when you need to use a previous version of a container image.