Create recommendation-for-testing.md

playbooks/ml-for-testing.md

@@ -0,0 +1,53 @@
+# Setting up Machine Learning jobs for testing
+
+Start with a folder or repo that has all the code *plus* a Dockerfile to help with running on AWS. Look at this [Dockerfile](https://github.com/artsy/entropy/blob/main/entropy/homefeed-artwork-reco/Dockerfile) for inspiration.
+
+## Setting up AWS ECR
+- In the [ECR dashboard](https://console.aws.amazon.com/ecr/repositories), you can create a new *private* repository by clicking the "Create repository" button on the top right.
+- Use a descriptive name
+
+After it's created, click the "View push commands" button on the top right to see the commands needed to run locally to upload a built docker image to ECR.
+The steps are basically:
+- get a login password
+- build the docker image
+- tag the docker image
+- push the docker image
+
+After all that, you should have an image with the `latest` image tag in the newly created ECR repository. Click on it and make a note of the "Image URI", we will need it later.
+
+For more documentation, look [here](https://docs.aws.amazon.com/AmazonECR/latest/userguide/repository-create.html).
+
+## Setting up AWS Batch
+- Open the AWS Batch console first-run wizard at https://console.aws.amazon.com/batch/home#/wizard
+- Set up the environment in step 1
+  - Use Fargate for provisioning model
+  - Set maximum vCPUs if needed
+  - Make sure you select the right networking VPC ID (use production)
+- Set up the job queue in step 2
+  - Priorty 1 is the default, change it if needed
+- Set up the job definition in step 3
+  - Set the image name in the container properties to be the `latest` Image URI from the ECR step above
+  - Set the command (usually `python main.py`)
+  - Set vCpus to maximum
+  - Set memory to maximum
+  - Execution role should be `artsy-data-infrastructure-iam-core-task-exc-production`
+  - Set "Assign public IP" to Enable (optional)
+  - Just above Parameters, click on "Additional configuration" to set the envvars under "Environment variables configuration". (`DB_REDSHIFT_NAME`, etc)
+- Set up the first job in step 4
+  - Set the command, vCpus and Memory to whatever you used in the previous step
+- Review in step 5 and you're done!
+
+For more documentation, look [here](https://docs.aws.amazon.com/batch/latest/userguide/Batch_GetStarted.html).
+
+## Setting up AWS EventBridge
+In the [EventBridge dashboard](https://console.aws.amazon.com/events/home?region=us-east-1#/rules) you can create a new rule that will schedule job for you
+- Click the "Create rule" button on the top right
+- You can use the "Schedule" pattern, with either cron or a fixed rate
+- Set the target to "Batch job queue"
+- Fill in the form with the ARNs
+  - Find "Job queue" in Batch. Click on "Job queues" and then click on your new job queue, and find the ARN.
+  - Find "Job definition" in Batch. Click on "Job definitions" and then click on your new job definition, and find the ARN.
+  - Enter a job name, that will be the name that each new job gets.
+  - Create a new role for this specific resource
+
+After all this, you should be able to see new jobs being created in the [ECR dashboard](https://console.aws.amazon.com/batch/home?region=us-east-1#dashboard)! Congrats.